<%BANNER%>

Cross-National Study of Childrens Temperament Structural Validity of the Student Styles Questionnaire

Permanent Link: http://ufdc.ufl.edu/UFE0044947/00001

Material Information

Title: Cross-National Study of Childrens Temperament Structural Validity of the Student Styles Questionnaire
Physical Description: 1 online resource (157 p.)
Language: english
Creator: Callueng, Carmelo M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: temperament -- validity
Special Education, School Psychology and Early Childhood Studies -- Dissertations, Academic -- UF
Genre: School Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Temperament has a long history of scholarship dating back as early as 350 BC when Hippocrates (1984) associated body fluids or temperament with behavior. Jung’s ideas contribute to modern temperament theory. Temperament is broadly described as stylistic and relatively stable traits that subsume intrinsic tendencies to act and react in somewhat predictable ways to people, events, and other stimuli. Advances in temperament research include the development of scientific and practical measures that can be used to obtain information to better understand a person’s social-emotional and behavioral functioning. The Student Styles Questionnaire (SSQ) is a measure of temperament originally developed for use with children in the U.S. and is now adapted in more than 30 countries. Psychometric characteristics of the SSQ adapted versions are scanty. To this end, this research examined the cross-national stability of a multidimensional model of temperament comprising of extroverted-introverted, practical-imaginative, thinking-feeling, and organized-flexible traits. Using the item level factor analytic framework with dichotomous items as indicators, results indicated a poor fit of the four-factor model of temperament in all the 21 countries. Not all item indicators converged to the traits they intended to measure. High correlations are evident in data 8 for some countries especially between the practical-imaginative and organized-flexible traits. Moreover, an alternative factor models are explored with the data from China, Egypt, and Gaza as a result of multicollinearity of factors. Findings are discussed in light of the shortcomings of factor analysis using non-normal dichotomous items and examining construct validity of adapted measures. Alternative approaches are suggested to further evaluate the factor structure of temperament as measured by the Student Styles Questionnaire.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Carmelo M Callueng.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Oakland, Thomas D.
Local: Co-adviser: Joyce, Diana.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044947:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044947/00001

Material Information

Title: Cross-National Study of Childrens Temperament Structural Validity of the Student Styles Questionnaire
Physical Description: 1 online resource (157 p.)
Language: english
Creator: Callueng, Carmelo M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: temperament -- validity
Special Education, School Psychology and Early Childhood Studies -- Dissertations, Academic -- UF
Genre: School Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Temperament has a long history of scholarship dating back as early as 350 BC when Hippocrates (1984) associated body fluids or temperament with behavior. Jung’s ideas contribute to modern temperament theory. Temperament is broadly described as stylistic and relatively stable traits that subsume intrinsic tendencies to act and react in somewhat predictable ways to people, events, and other stimuli. Advances in temperament research include the development of scientific and practical measures that can be used to obtain information to better understand a person’s social-emotional and behavioral functioning. The Student Styles Questionnaire (SSQ) is a measure of temperament originally developed for use with children in the U.S. and is now adapted in more than 30 countries. Psychometric characteristics of the SSQ adapted versions are scanty. To this end, this research examined the cross-national stability of a multidimensional model of temperament comprising of extroverted-introverted, practical-imaginative, thinking-feeling, and organized-flexible traits. Using the item level factor analytic framework with dichotomous items as indicators, results indicated a poor fit of the four-factor model of temperament in all the 21 countries. Not all item indicators converged to the traits they intended to measure. High correlations are evident in data 8 for some countries especially between the practical-imaginative and organized-flexible traits. Moreover, an alternative factor models are explored with the data from China, Egypt, and Gaza as a result of multicollinearity of factors. Findings are discussed in light of the shortcomings of factor analysis using non-normal dichotomous items and examining construct validity of adapted measures. Alternative approaches are suggested to further evaluate the factor structure of temperament as measured by the Student Styles Questionnaire.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Carmelo M Callueng.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Oakland, Thomas D.
Local: Co-adviser: Joyce, Diana.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044947:00001


This item has the following downloads:


Full Text

PAGE 1

1 CROSS STRUCTURAL VALIDITY OF THE STUDENT STYLES QUESTIONNAIRE By CARMELO M. CALLUENG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFIL LMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Carmelo M. Callueng

PAGE 3

3 ACKNOWLEDGMENTS I would like to express my sincerest appreciation to the significant persons who immensely co ntributed to the completion of my dissertation. I thank my advisor and committee chair, Dr. Thomas Oakland, for his competent advising and unwavering support for this project and throughout my graduate studies at the University of Florida. His expertise in temperament and international psychology has strongly motivated me in the pursuit of this project. His dedication to hard work, excellence, scholarly rigor, and research has inspired me to enhance my academic competence and professional commitment. On a p I thank my co chair, Dr. Diana Joyce, for her guidance and reliability. I am indebted to her for t he countless assistance, positive regard, and encouragement in getting through this project and in my doctoral program. Her endearing qualities and values as a professor and as a person are worth emulating for. I also gratefully acknowledge my committee me mbers, Drs. John Kranzler and Walter Leite, for their commitment and contributions to this project. Their feedback, guidance, and suggestions were valuable in the fruition of this project. Lastly, I would like to express de epest gratitude to my family and acknowledge the value of their unconditional love and support. I am sincerely grateful to my mother, Francisca Callueng, my late father, Domingo Callueng, and my siblings (Teddy, Roderick, Joyce, Rudolph, and Clifford) and their families.

PAGE 4

4 TABLE OF CONT ENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 6 ABSTRACT ................................ ................................ ................................ ..................... 8 CHAPTER 1 REVIEW OF LITERATURE ................................ ................................ .................... 10 Theoretical Perspective of Temperament ................................ ............................... 12 Classic Theory of Temperament ................................ ................................ ...... 12 Modern Theory of Tempe rament ................................ ................................ ...... 14 Temperament Construct in Children and Youth ................................ ...................... 21 Thomas and Chess Temperament Theory ................................ ....................... 22 Psychological Type Approach to Temperament ................................ ............... 23 Developmental Perspective of Temperament ................................ ......................... 29 Temper ament Preferences of Children in Special Populations ............................... 30 Children with Anxiety Disorder and Depressive Disorder ................................ 30 Children with Oppositional Defiant Disorder (ODD) and Conduct Disorder (CD) ................................ ................................ ................................ .............. 31 Children with Attention Deficit Hyperactivity Disorder (ADHD) ......................... 32 Children with Autism Spectrum Disorder (ASD) ................................ ............... 33 Children with Learning Disabilities ................................ ................................ .... 34 Children Identified as Gifted ................................ ................................ ............. 34 Children with Visual Impairments ................................ ................................ ..... 35 Correlation between Temperament and Cognitive Variables ................................ .. 35 Cross national Differences in Temperament in Children ................................ ........ 36 Test Validity ................................ ................................ ................................ ............ 39 Research Objectives and Hypot heses ................................ ................................ .... 42 2 METHODS ................................ ................................ ................................ .............. 43 Participants ................................ ................................ ................................ ............. 43 Measure ................................ ................................ ................................ .................. 47 Procedure ................................ ................................ ................................ ............... 51 Data Analyses ................................ ................................ ................................ ......... 53 Data Screening ................................ ................................ ................................ 53 Internal Consistency of the SSQ ................................ ................................ ...... 53 Structural Validity ................................ ................................ ............................. 54 3 RESULTS ................................ ................................ ................................ ............... 60 Data Screening ................................ ................................ ................................ ....... 60

PAGE 5

5 Internal Consistency ................................ ................................ ............................... 60 Structural Validity ................................ ................................ ................................ .... 60 4 DISCUSSION ................................ ................................ ................................ ....... 134 Implications ................................ ................................ ................................ ........... 140 International Research ................................ ................................ ................... 1 40 Structural and Measurement Equivalence ................................ ...................... 141 Test Validity ................................ ................................ ................................ .... 142 Limitations ................................ ................................ ................................ ............. 143 Conclusions ................................ ................................ ................................ .......... 144 LIST OF REFERENCES ................................ ................................ ............................. 146 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 156

PAGE 6

6 LIST OF TABLES Table page 2 1 Sample sizes of countries by total, gender, and age group ................................ 58 2 2 SSQ translation languages, names and institutional affiliations of on site researchers ................................ ................................ ................................ ......... 59 3 1 Participants per country with at least 6 item responses missing ......................... 91 3 2 Cronbach alpha coefficients of the temperament traits in 21 countries .............. 92 3 3 Overall goodness of fit indices for individual coun try CFA ................................ 93 3 4 Parameter estimates of the four factor model in Australia data .......................... 95 3 5 Parameter estimates of the fo ur factor model in Brazil data ............................... 97 3 6 factor solution for China data ................................ ................................ .............. 99 3 7 Parameter estimates of the four factor model in Costa Rica data .................... 100 3 8 factor solution for Egypt data ................................ ................................ ............ 102 3 9 factor solution for Gaza data ................................ ................................ ............ 103 3 10 Parameter estimates of the four factor model in Hungary data ........................ 104 3 11 Parameter estimates of the four factor model in Iran data ................................ 106 3 12 Parameter estimates of the four factor model in Israel data ............................. 108 3 13 Parameter estimates of the four factor model in Japan data ............................ 110 3 14 Parameter estimates of the four factor model in Mongolia data ....................... 112 3 15 Parameter estimates of the four factor model in Nigeria data .......................... 114 3 16 Parameter estimates of the four factor model in Pakistan data ........................ 116 3 17 Parameter estimates of the four factor model in Philippines data ..................... 118 3 18 Parameter estimates of the four factor model in Poland data ........................... 120 3 19 Parameter estimates of the four factor model in Romania data ....................... 122

PAGE 7

7 3 20 Parameter estimates of the four factor model in Samoa data .......................... 124 3 21 Parameter estimates of the four factor model in Singapor e data ...................... 126 3 22 Parameter estimates of the four factor model in U.S. data ............................... 128 3 23 Parameter estimates of the modi fied four factor model in Venezuela data ...... 130 3 24 Parameter estimates of the modified four factor model in Zimbabwe data ....... 132

PAGE 8

8 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CROSS STRUCTURAL VALIDITY OF THE STUDENT STYLES QUEST IONNAIRE By Carmelo M. Callueng December 2012 Chair: Thomas Oakland Cochair: Diana Joyce Major: School Psychology Temperament has a long history of scholarship dating back as early as 350 BC when Hippocrates (1984) associated body fluids or temperame nt with behavior. Temperament is broadly described as stylistic and relatively stable traits that subsume intrinsic tendencies to act and react in somewhat predictable ways to people, events, and other stimuli. Advances in temperament research include the development of scientific and practical measures that can be used to obtain information to better emotional and behavioral functioning. The Student Styles Questionnaire (SSQ) is a measure of temperament based on the type theory of Jung, and Myers and Briggs. The SSQ was originally developed for use with children in the U.S. and is now adapted in more than 30 countries. Few studies have reported the psychometric characteristics of the SSQ adapted versions. To this end, this resea rch examined the cross national factor structure of temperament comprising of extroverted introverted, practical imaginative, thinking feeling, and organized flexible traits. Using the item level factor analytic framework with dichotomous items as indicato rs, results indicated that the four factor model of temperament did not fit reasonably well to the

PAGE 9

9 data in each of the 21 countries. M ost item indicators converged on the traits they intended to measure. However, factor loadings generally were low. In most countries, the bipolar traits are distinct on what they measure. However, multicollinearity was evident in data from China, Egypt, and Gaza. Findings are discussed in light of the shortcomings of factor analysis using dichotomous items when examining cons truct validity of the SSQ and sample size requirement s of robust weighted least square estimator in CFA Implications and limitations of the study also are discussed.

PAGE 10

10 CHAPTER 1 REVIEW OF LITERATURE Scholarship on temperament has had an important agenda that includes its conceptual meaning, measurement, and establishing a network of variables that are believed to be associated with temperament. Temperament currently is thought to characterize stylistic and relatively stable traits that subsume intrinsic t endencies to act and react in somewhat predictable ways to people, events, and other stimuli (Teglasi, 1998a; 1998b). A number of temperament theories present divergent views, resulting to a lack of consensus definition. Scholars differ on their views on what behaviors constitute temperament and the number of dimensions or traits that constitute temperament. Temperament can encompass behavioral styles, emotions, degrees of stability, and heritability (Goldsmith, et al., 1987). Outweighing these areas of divergence are unifying views that serve as foundational ideas in understanding the co ncept of temperament (Goldsmith et. al., 1987). Scholarship generally supports the belief that temperament constitutes a group of related traits rather than a single tra it, addresses the issue of individual differences rather than specifies general characteristics, has biological origin, adheres to continuity in behavioral manifestations, is dynamic despite adherence to behavioral continuity, and refers to specific moods, attitud es, and dispositions that affect behaviors. temperament in later life (Wach, 1994). Oakl and, Glutting, and Horton (1996) expanded the temperament perspective in children by applying the Jungian typology and the Myers/Briggs type theory in the development of the Student Styles Questionnaire (SSQ)

PAGE 11

11 for children ages 8 through 17. The SSQ provide s information on temperament preferences in four bipolar traits: extroverted introverted, practical imaginative, thinking feeling, and organized flexible. of the origins of behav iors and attitudes children display at home and school and its normal development. Temperament can assist in identifying talent, adjusting for possible weaknesses, enhancing personal and social development, promoting an understanding of others, assessing l earning styles, promoting educational development, exploring career interest, and facilitating research and evaluation studies (Oakland, Glutting, & Horton, 1996). These broad uses of temperament are closely aligned to an ultimate goal of school psychology namely to provide comprehensive and multilevel psychological services to children and youth that enable them to succeed academically, socially, behaviorally, and emotionally. fr om cross national perspectives has increased recently (Buss, 2011). This surge of interest in cross national psychology over the past two decades with the goal to develop a broadly shared framework that recognizes the bidirectional roles of person and the environment in understanding behavior. Specifically, temperament research can be viewed from the lens of cultural and global diversity by engaging in studies that focus on theoretical and current practices in child and adolescent development. The current study recognizes the value of valid measures to obtain information that can be used to make decisions relating to adjustment of children and youth. It primarily contributes in understanding the meaning of temperament cross nationally and

PAGE 12

12 explores the emerg ing view that the development of temperament may be tempered by the environment and personal choice through the process of enculturation and accommodation, an idea closely related to psychological anthropology that views culture as constitutive of behavior (Miller, 1999). Few studies examined the cross national structural validity of the Student Styles Questionnaire (SSQ) as a measure of temperament in children. This study addresses the psychological meaning of temperament by examining its construct equival ence using data from children in 21 countries. Theoretical Perspective of Temperament The concept of temperament has evolved from a simple idea related to body fluids to a more sophisticated and broad theoretical explanation that encompasses biological and environmental influences. The following discussion presents theoretical and historical perspective of temperament. Classic Theory of Temperament Hippocrates Temperament was introduced in the Greek culture as early as the 4 th century B.C. by Hippocrates Temperament was described as behavior that clusters in four humors, namely yellow and black bile, blood, and phlegm. They were considered to be neurochemical precursors of bipolar behaviors thought to oppose each other (i.e., warm and cool; dry and moist) on the premise that all mental, emotional, and behavioral disorders are caused by natural factors such as inherited susceptibility and an imbalance of the four humors (Hergenhahn, 2001). Galen Galen expanded on physical and emotional characteristics of temperaments (Galen, trans, 1992;

PAGE 13

13 Hergenhahn, 2001; Hippocrates, trans. 1939): choleric, phlegmatic, melancholic, and sanguine. Galen believed that the emotional an d behavioral functioning was due to one pair of bodily states dominating the complementary pair (e.g., warm and moist dominating cool and dry). The choleric type has an over supply of yellow bile that would incline a person to be both easily angered and ea sily calmed, to quickly change moods and likes, to be irascible, and, in its extreme, to be seen as a fool. The phlegmatic type has an over supply of phlegm that would incline a person to be pale, slow, drowsy, apathetic, weak, to engage in fantasy, be mil d mannered, and exhibit somatic complaints. T he melancholic type has an over supply of black bile that would incline a person to display extremes of happiness or malaise, sadness, and depre ssion. The sanguine has an over supply of blood that would incline a person to be pleasant, loving, affectionate, happy, optimistic, and hopeful (Galen, trans, 1992; Hergenhahn, 2001) Kant In 1798, Kant published Anthropologie which included a c hapter on temperament. He discussed how anthropology can become a viable alt ernative to Kant believed that anthropology can provide sufficient informational cues to predict and s were the basic elements of the four distinct temperaments: choleric, phlegmatic, melancholic, and sanguine. Persons who display a choleric temperament type generally are easily irritated and become upset when others do not listen to what they say. Person s who display a phlegmatic temperament type generally are persistent and use reason rather than instinct to guide their actions. Persons who display a melancholic temperament type generally are fearful and sad. Persons who display a sanguine temperament ty pe

PAGE 14

14 generally are carefree and sociable. In terms of behavioral characteristics, Kant further (sanguine vs. melancholic). A person can have a distinct temperament based on his or her dominant behavioral characteristics. However, a combination of temperament types (choleric phlegmatic) does not exist. Modern Theory of Temperament Wundt, known as the father of experim ental psychology, introduced a somewhat m odern view of temperament in 1903 that closely resembled the four temperaments proposed by Galen and Kant (Eysenck & Eysenck, 1985) temperament, Wundt believed that temperament can be measured and quantified and that a person can display one or a combination of positions on the energetic and emotional temperament dimensions. Wundt used the terms strong emotions in combination with weak emotions, and changeable in combination with unchangeable heory observation and interaction with clinic patients. He noted that patients with similar temperament qualitie s displayed similar adjustment problems. For example, he reported that patients with hysteria displayed extroverted attitudes regardless of their emotional instability. He described hysteric patients as being consistently aware of their surroundings and in teracting actively with the therapist. On the other hand, he reported that patients with schizophrenia displayed introverted attitudes and preferred to be alone and withdrawn from others (Storr, 1991).

PAGE 15

15 While acknowledging the ideas of Hippocrates and Gal en about temperament, Jung proposed a differentiated temperament theory of psychological typology. Unlike Galen who believed that temperament is based on emotions or feelings, Jung explored the unconscious mind through the study of attitudes and values. Ju ng described attitude 414). The introversion and extroversion. He believed that every pe rson displays both introversion and extroversion tendencies and that a person is not entirely introverted or extroverted in attitude (Jung, 1921/1971). However, each person is thought to have a particular preference for introverted or extroverted attitudes and most often display behaviors consistent with their preferred attitude. Introversion and extroversion also can be understood as opposing attitudes along a continuum. Thus, persons may vary from strongly introverted to slightly introverted or strongly e xtroverted to slightly extroverted (Jung, 1921/1971). Jung also noted that extroversion and introversion attitudes bring 1971). An introvert is drawn to their own thoug hts and inner feelings. They prefer to be alone or in small groups, are introspective, tend to be cautious, and make decisions slowly. In contrast, extroverts are drawn to their outside environment. They are sociable, xpectations (Jung 1921/1971). In addition to the introversion and extroversion, Jung identified four basic psychological functions: thinking, feeling, sensation, and intuition (Jung, 1921/1971). Each of these functions can be displayed in tandem with intr oversion and introversion.

PAGE 16

16 Jung (1920/1926) believed that persons tend to have well developed and dominant functions that are used more on a conscious level. There also are secondary functions that are not well developed and typically are used on a conscio us level, yet can be potentially developed if desired by a person. The functions operate in pairs on a conscious level at any given time (Jung, 1921/1971). Jung viewed thinking and feeling as rational functions (Jung, 1921/1971). Persons who display a thin king function make decisions through careful deliberation together with the use of logical and objective information. They highly value justice, truth, and facts. Persons who display a feeling function make decisions through a s personal values (e.g., loyalty, sympathy, harmony with others). This value creates a sense of liking, disliking, or general mood that integrates kind of judgment, diff ering from intellectual judgment in that its aim is not to establish (Jung, 1921/1971, p. 434). According to Jung (1921/1971), feeling is a rational quality because the l aws of reason are used in establishing value. Jung believed intuition and sensation dimensions represent two different functions for acquiring and assessing information. Intuition and sensation operate in opposite directions and are considered irrational d ecision making styles. Persons who display an intuition function assimilate quickly, thus integrating past experiences and unconscious perceptions. In contrast, persons who display a sensation function require direct physical experiences of the world, with facts or external stimuli acquired through

PAGE 17

17 the physical senses. Unlike intuition, sensation is a conscious perception and is governed more predominantly by concrete experience than by analysis. The pairing of the combination of temperament functions resu lted in eight temperament types. Jung believed that a person displays a dominant and frequently used type and an auxiliary or secondary type that is used less frequently (Jung, 1921/1971). He referred to four of these types as rational and the other four a s irrational. The rational types are extroversion thinking, introversion thinking, extroversion feeling, and introversion feeling. Jung identified famous persons who exemplify the qualities of each rational type. Charles Darwin, known for his penchant for scientific reasoning and facts, exemplifies an extroversion thinking type. Immanuel Kant, known for the importance he placed on subjective reality and rational thinking, exemplifies an introversion thinking type. Although both types are strongly influence d by ideas, the extroversion thinking type is interested in objective data and will follow ideas externally. The introversion thinking type is influenced by subjective ideas and will ponder those inwardly (Jung, 1921/1971). Jung believed that the extrovers ion feeling and introversion feeling types commonly are displayed in women. These types are guided by personal value systems comprised of subjective feelings and place strong value on harmony. intuitive, introv ersion intuitive, extroversion sensing, and introversion sensing. His example of an introversion intuitive type is a person who is a dreamer or artist who enjoys contemplative moments. His example of an extroversion intuitive type is a person who exhibits strong dependence on the external environment and is searching for new possibilities. Each of these types

PAGE 18

18 is strongly influenced by subjective factors and ideas. Compared to an extroversion sensing type who seeks external facts, concrete objects, and ideas an introversion sensing type analyzes or thinks deeply on ideas and information (Jung 1921/1971). Jung discussed the phenomenon of falsification of type in relation to his theory of psychological type. Falsification of type occurs when a person uses his or her non preferred functions more than his or her highly efficient and natural preferred functions. This phenomenon is caused when a person is placed in an environment that lessens or impedes opportunities to fully allow the expression of his or her nat ural preference. On the other hand, non falsification of type occurs when a person is placed in an environment that allows the expression of his or her natural preference, resulting in a positive psychological well being. A favorable adjustment is achieved when a person experiences a close match between his or her strengths and the demands of the environment (Jung 1921/1971). This belief is consistent with Thomas and (1977) notion of goodness of fit. Myers and Briggs Theory Myers was fascinated wi fourth dimension, judging and perceiving. Judging or perceiving are concepts that relate to how persons prefer to structure their lives in relation to the environment (Myers & Myers, 1980). A person with a judging orientation has a preference for planning, systems, order, routine, standards, self regimentation, purposeful actions, decisiveness, and closure. A person with a perceiving orientation has a preference for spontaneity, understanding, tolerance, curiosity, zest for experi ence, and adaptability (Myers & Myers, 1980).

PAGE 19

19 psychological types. The Myers Briggs Type Indicator (MBTI) was published 20 years later (Myers & Myers, 1980). The MBTI combines Jun dimensions [extroversion (E) introversion (I), sensing (S) intuition (N), and thinking (T) perceiving (P) dimension. The extroversion rientation. A person who prefers an extroverted orientation draws energy from the outer world of peopl e and events while a person who prefers an introverted orientati on draws energy from the self, including thoughts and introspection (Joyce, 2010). The se nsing processes. A person who prefers a sensing function acquire information from the physical senses, is realistic, and oriented to details. A person who prefers an intuitive function acquire s information through logical and theoretical deductions, lesser attention to details, and more analytical in processing information (Joyce, 2010). The thinking making process. A person who prefers a thinkin g function make decisions based on facts, logic, and objective data. Fairness is achieved by applying the principles of justice and truth. A person who prefers a feeling function make decisions based on subjective values, such as empathy and well being of others (Joyce, 2010). The judging orientation. A person who prefers a judging orientation values structure when interacting with the outside environment, is organized, and plans ahead. A pe rson who prefers a

PAGE 20

20 perceiving orientation interacts with the outside environment is a flexible manner, is adaptive, and likes to keep options open (Joyce, 2010). The four dimensions yield 16 type combinations: ISTJ, ISFJ, INFJ, INTJ, ISTP, ISFP, INFP, INT P, ESTP, ESFP, ENFP, ENTP, ESTJ, ESFJ, ENFJ, and ENTJ Guidelines for administration, scoring, and interpretation of the MBTI are described in the Essentials of Myers Briggs Type Indicator Assessment, Second Edition (Quenk, 2009). As a measure of temperame nt or personality, the MBTI was conceptualized as a tool to obtain information that aids in understanding individual differences rather than a measure of pathology (Myers & Myers, 1980). Today, the MBTI is one of the most widely used personality test by ps ychologists, counselors, social workers, and other mental health professionals in the world. Industrial/Organizational psychologists and human resource specialists also utilize the MBTI for career assessment, employee training, and team building programs ( Joyce, 2010). Keirsey Theory The Keisey Temperament Sorter provides a brief, self scoring, temperament measure that yields the MBTI 16 types (Keirsey & Bates, 1978). However, he introduced a modified interpretation of the 16 types into fo ur clusters of interpretation based on the work of other theorists (i.e., Ernst Kretschmer, Eduard Spranger, Eric Adickes, and Eric Fromm). The four clusters included sensing judging, sensing perceiving, intuition thinking, and intuition feeling. The sens ing judging temperament is characterized as responsible, conservative, stable, productive, organized, and a strong work ethic. Persons who prefer this temperament type are compelled to fulfill obligations, are industrious, have a high need

PAGE 21

21 for belongingnes s, and often are caretakers. The sensing perceiving temperament is characterized as active, spontaneous, open minded, and adaptable. Persons who display this temperament are cheerful and enjoy exploration and discovery. They have a high need for freedom an d have a strong play ethic (Keirsey & Bates, 1978). The intuition thinking temperament is characterized as rational, analytical, systematic, and research oriented. Persons who display this temperament generally exhibit high standards, are curious, displa y a high need for achievement, are perfectionistic, and may be compulsive. The core value of an intuitive thinking person is to develop competencies and skills. They put work before play and can incorporate play into work by striving to develop recreation skills, such as golfing expertise. The intuitive feeling temperament is characterized as friendly, imaginative, caring, and sensitive to the needs of others. Persons who display this temperament are non competitive and passionate about social causes and th e impact of their actions on humanity. The core value of intuitive feeling person is personal integrity and self actualization (Keirsey & Bates, 1978). Temperament Construct in Children and Youth Early theory and research on temperament were conceptualize d on adult behavior patterns. However, theorists agree that temperament dispositions are early appearing and that temperament differences are present as early as infancy, thus suggesting temperament is innate or biologically rooted (Jung, 1928/1945; Rothba rt, to the environment, especially how quick they interact with objects and o ther people, as

PAGE 22

22 an early sign of extroversion. Indicators of introversion in children include shyness, reflective thinking, and fearfulness of unknown objects. Thomas and Chess Temperament Theory Thomas and Chess (1977, 1984, 1986, & 1989) developed a fram ework for explained the similarity between behavioral and temperament style. Behavioral style describes how a person behaves, rather than how well, what (i.e., abilities and co ntent), or why (i.e., motivation) a person behaves. Moreover, behavioral style refers to behavioral characteristics that are present at birth through later life. In contrast, temperament refers to behavioral tendencies that appear in early infancy. Thomas and Chess (1977, 1984, 1986, & 1989) considered development as a complex process of interplay between the child and the environment. They believed that temperament is best understood when it is explained in the context of the environment. They used the co ncept of goodness of fit to explain the temperament environment interactive process. Goodness of fit is achieved when there is a match between the of fit is the consequ ence of a mismatch between the two sets of desired characteristics. Thomas and Chess (1977, 1984, 1986, & 1989) described nine behavioral categories of temperaments based on data obtained though observation, parent questionnaires, and teacher interviews r egarding the infancy periods of 22 children. temperament patterns: easy, difficult, and slow to warm up. Easy temperament was displayed by approximately 40% of the infants. They were described as being able to establish regular routines, cheerful, and able to adapt easily to new situations. Parents

PAGE 23

23 further described these children as contented and easy going. A difficult temperament was displayed by approximately 10% of the infants. About 70% of these children were expected to encounter long term adjustment problems at some point in the future. They were described as experiencing irregular routines, were slow to adapt to new circumstances, displayed problematic sleep cycles, and tended to react negatively. Slow to warm up temperament was displayed by approximately 15% of the infants. They were described as watchful with strangers, lethargic, tending to display a negative mood, and adjust rather slow to new situations. Approximately 35% of the infants displayed a combination of temperament patterns. Psychological Type Approach to Temperament The psychological type approach was adopted by the SSQ to understand temperament in eight basic styles and grouped into four bipolar traits, namely extroverted or introverted, practical or imaginative, thinking or feeling, and organized or flexible. In turn, the basic styles can form 16 meaningful comb inations with four styles (e.g., extroverted practical thinking organized) per combination (Oakland, Glutting, & Horton, 1996). Extroverted introverted traits Extroverted introverted traits are associated with the source from which a student derives his o r her energy. Approximately 65% of U.S. children prefer extroverted style and 35% prefer introverted style (Oakland, Glutting, & Horton, 1996). Children who display a preference for extroverted style derive energy from the external environment and are more oriented to people and events. On the other hand, children who prefer an introverted style derive energy from themselves or their immediate family and close friends and are more oriented to ideas and reflection

PAGE 24

24 Children who prefer an extroverted style e njoy meeting and interacting with people. They exhibit high energy to socialize with people and are open to discussions about different topics. Extroverted children are drawn to participate in group activities and would feel uncomfortable when required to spend too much time alone. They are They develop their thoughts and ideas through oral discussion and thrive in a learning environment where there are ample opportunities to engage in hands on activities. Positive behaviors of extroverted students are reinforced by encouragement and praises. Discussion is the primary means to develop their thoughts and ideas. Extroverted children may not be comfortable to work on long assign ments or seatwork. They also tend to act without thinking and planning. Their penchant for interaction may disrupt the classroom environment and when extreme, may be annoying to students who display strong preference for introverted style. In contrast, chi ldren who prefer an introverted style prefer to be alone and may feel exhausted and bored with frequent and long interaction with others. They are slow in responding to their surrounding and have a wait and see tendency before joining an activity. Introver ted children are cautious, have fewer friends, and prefer to listen than talk in a group. They can work for long periods on a project without interruption. Introverted children tend to be selective of people they want to be close with. Their reserved and quiet manner sometimes is misinterpreted as displaying poor social skills. In school, they prefer to work with other children who match their behavioral qualities in small groups or in pairs. They like doing tasks that require methodical, reflective, and r easoning skills. They prefer to be called by the teacher to recite in class

PAGE 25

25 rather than volunteer to recite. At times, teachers and other adults in school would label introverted students as unfriendly, uncooperative, and withdrawn because of their prefere nce to be alone and quiet. Practical imaginative traits Practical imaginative traits are associated with what students attend to. Approximately 65% of U.S. children prefer a practical style and 35% prefer imaginative style (Oakland, Glutting, & Horton, 19 96) Children who prefer a practical style are more realistic and pragmatic in their view of the world. They tend to pay more attention to details and facts when acquiring information. Practical children are keen observers and use their physical senses to know more about their environment. They enjoy family traditions, recreation, and leisure time. They find the company of their extended family members satisfying. Practical children enjoy learning through a step by step approach. They are organized, have g ood rote memory, and persist in working toward their goals. For them, the present is more important and they prefer to do things that have practical value in their lives. Practical children are more concrete learners and thus, become disinterested in learn ing complex material and abstract concepts. In contrast, children who prefer an imaginative style enjoy activities that harness their creative skills and qualities. They also enjoy acquiring new skills and do things in non traditional ways. They love ideas and concepts rather than factual and sensory knowledge. Imaginative students are more engaged in tasks that require analysis, comprehension, and formulating conclusions. They prefer to learn the theory first and proceed to know its usefulness and applicat ion. Imaginative children are challenged by difficult situations and may confidently take risks.

PAGE 26

26 Imaginative children value quickness of understanding, especially in learning novel tasks. They enjoy testing hypotheses and generating insights about their e xperiences. Within the family, imaginative children generally have close attachments to their parents, siblings, and the extended family members. In school, they prefer to work on projects and assignments that require imagination and analysis. They are fas cinated to learn new theories, even trivial ones that lack empirical evidence. At times, they come up with very complex and sophisticated plans that can be difficult to implement. Thinking feeling traits Thinking feeling traits are associated with how chi ldren make decisions. These traits are significantly influenced by gender. Approximately 65% of U.S. males and 35% of U.S. females prefer a thinking style. On the other hand, approximately 65% of U.S. females and 35% of U.S. males prefer a feeling style (O akland, Glutting, & Horton, 1996) Children who prefer a thinking style are likely to be fair and objective in making decisions. They attach more importance to logic than emotions. Thinking children are more engaged in activities that require analysis of facts and ideas. They are more likely to exhibit a skeptical stance on what other says and typically ask questions to challenge In social situations, children who prefer thinking style are likely to enjoy the company of peers who share common interests. When with friends, they prefer to do tasks that stimulate their logical and analytical skills rather than engage in social conversation and other more socially oriented activities. They are open with their opinions that may not be supported by evidence. Their tendency to express criticisms may be offending to others. In the family, students who prefer thinking style tend to be

PAGE 27

27 close to their parents and other family members. However, they are less likely to express their love and affection, both verbally and non verbally. They cooperate and follow family rules that seem fair and reasonable. As students, they are more likely to sustain attention during lectures and presentation that are organiz ed logically. They are more likely to enjoy subjects (e.g., math and science) that require the use of logical and analytical skills. Feedback to them is beneficial when teachers point out their errors and how to correct them. In contrast, children who p refer a feeling style rely on their emotions and interpersonal relationship when making decisions. They are more engaged with people, hence are friendly, charming, sympathetic, and generous with appreciation. They support or participate in projects that pr omote human welfare, peace, and harmony. In social relationships, children who prefer feeling style tend to avoid conflict or disagreement with others that may strain relationships. They give more importance to friendships and are careful not to hurt feeli ngs of others. In the family, children who prefer feeling style tend to be very expressive with their love and affection to parents and other family member. They are likely to please their parents by doing chores and following rules at home. They are carin g, empathetic, and supportive. Conflicts or disagreements in the family cause them distress. As students, they tend to enjoy subjects or lessons that deal with people. Praises and other forms of affirmation are effective strategies to motivate them to perf orm well. In group activities, they enjoy working with friends and may even accept more responsibilities to fill in for others. They shy away from activities that promote competition rather than cooperation.

PAGE 28

28 Organized flexible traits Organized flexible tr aits are associated with a tendency to either make decisions rather quickly or to postpone them. Approximately equal proportions of U.S. children prefer organized and flexible styles (Oakland, Glutting, & Horton, 1996) Children who prefer an organized st yle are likely to have structure and plan their activities. They have an admirable work ethic; are persistent, dependable, and exhibit self discipline. They generally complete tasks in a systematic way. Organized students generally show respect to authori ty. In social relationships, children who prefer organized style generally want their friends to conform to their expectations. Loyalty and support to friends are defining qualities. They tend to devote much time to prepare for social activities and may be upset when changes are made on the plans, especially at the last minute. In the family, they enjoy a well planned home life and expect parents and other family members to fulfill their promises. They generally are obedient to rules and do chores in an ord erly way. As students, they generally have good study habits and are in control of their school activities. They are organized in their work and devote ample time to prepare for their activities. Teachers and other adults in school perceive students who d isplay proclivity for organized style as responsible, obedient, and dependable. They expect teachers to provide explicit instructions and guidelines related to grading, examinations, assignments, and classroom behavior. They generally put forth their best effort to attain goals and usually are motivated to maintain good work and behavior through consistent praise and rewards.

PAGE 29

29 In contrast, children who prefer a flexible style tend to adjust easily to changes in their lives and environment. They enjoy surpris es and may tend to avoid following rules. They are open in their commitments and simply want to take as many opportunities as possible to enjoy life. They are likely to be accepting of diversity and tolerant of opinions from others. In social relationships children who are inclined to flexible style may be considered the clown of the group because they enjoy making people laugh and have fun. They seek fun most of the time and would have difficulty coping with stress and problems. In the family, children wh o prefer a flexible style enjoy fun activities with family members. They have a high need for autonomy and may resist rules at home. They may not keep their room or personal space organized or neat. As students, they are likely to enjoy lessons and activi ties if presented in a game like manner. They like to participate and perform in school activities, especially when there are incentives for winning. They thrive in classroom environments in which students are given options, have flexible deadlines in assi gnment and projects, can be permitted to move around the room, and are expected to follow only a few necessary rules. Developme ntal Perspective of Temperament The New York Longitudinal S tudy reported that temperament differences between males and females a ppear shortly after infancy and increase with age on the following qualities: adaptability, approach/withdrawal, activity, and sensory threshold (Chess & Thomas, 1991). During the period from four months to four years, males are more adaptable and approach ing than females. Between ages 8 to 12, males display higher levels of ac tivity and sensitivity (Maziade et al., 1986).

PAGE 30

30 Studies o f temperament style of U.S. children ages 8 17, confirm the pr esence of gender differences (Bassett & Oakland, 2009; Oakland, Glutting, & Horton, 1996) As noted previously, m ore males than fe males prefer thinking and flexible styles while more fe males than males prefer organize and feeling style s Gender differences on thinking feeling appear early, at least by age 8, are susta ined through adulthood, and may be universal (Hammer & Mitchell, 1996; Myers McCaulley, Quenk, & Hammer, 1998; Myers & McCaulley, 1985 ). Temperament Preferences of Children in Special Populations Temperament traits are important characteristics that can provide a non characteristics that require intervention (Joyce & Oakland, 2005). An understanding of temperament traits of children with special needs can contribute to building a m ore positive interactions and relationships between these children and their significant adults (e.g., their parents and teachers ) Likewise, psychologists and other mental health professionals can use temperament information when designing interventions a nd determining possible reinforcement strategies in behavioral modification plans for children. Few studies used the four bipolar traits to describe temperament of children with special needs. Children with Anxiety Disorder and Depressive Disorder The temp erament traits of 70 middle school children who met the criteria for eligibility as emotionally disturbed, with a primary diagnosis of either depression or anxiety, were examined (Jennings, 2005). No specific temperament style was associated with condition s of depression and anxiety. In general, children diagnosed with either depression or anxiety indicated stronger preference for extroversion than

PAGE 31

31 introversion, practical than imaginative, thinking than feeling, and organized than flexible styles. In light of the goodness of fit hypothesis ( Thomas & Chess, 1984), preferences for organized flexible styles in depressed and anxious children were examined in depression, children with anxiety were expected to prefer an organized style and a classroom setting that placed greater emphasis on structure and conformity to rules. Furthermore, children who displayed both depression and anxiety and who prefer a flexible style were expecte d to have a stronger preference for a classroom setting that placed lesser emphasis on structure and conformity to rules. Both hypotheses were disconfirmed. Children who were anxious did not differ from those who were depressed in their preferences for org anized style and classroom setting that placed greater emphasis on structure and conformity to rules. Moreover, preference for a flexible style was not associated with preference for a classroom setting that place lesser emphasis on structure and conformit y to rules. Children who either were anxious and depressed indicated a preference for a teacher who displayed with high emotional responsiveness regardless of classroom conformity (Jennings, 2005). Children with Oppositional Defiant Disorder (ODD) and Cond uct Disorder (CD) Temperament traits of 80 children and youth ages 8 through 17, and with a current diagnosis of either Oppositional Defiant Disorder (ODD) or Conduct Disorder (CD), were compared (Joyce & Oakland, 2005). Children and youth with CD and ODD differed only on practical imaginative styles; those with ODD displayed a stronger preference for practical style. Furthermore, those with ODD also displayed a greater preference for practical and thinking styles than do their general population peers.

PAGE 32

32 Kn owledge that children and youth with ODD prefer practical and thinking styles can be useful in developing treatment plans. Children with practical styles generally prefer simplicity, details, facts, and the concrete and sequential processing of ideas. A pr eference for a practical style among children and youth with ODD can be associated with their need for explicit behavioral expectations along with specific consequences (Joyce & Oakland, 2005). In addition, concrete and sequential communication of expectat behaviors (Barkley, 1997; Patterson, 1982). On the other hand, the stronger preference of children and youth with ODD for a thinking style can be considered a weakness since this can be associated with a negative attribution bias (Dodge & Newman, 1981). Consequently, children with ODD may tend to express hostility in the way they relate with people. Moreover, their inability to regulate the expression of their blunt and critical opini ons may increase their defiance and aggressive behaviors (Joyce & Oakland, 2005). Children with Attention Deficit Hyperactivity Disorder (ADHD) Temperament traits of 83 children, ages 8 through 12, with a current diagnosis of Attention Deficit Hyperactivi ty Disorder (ADHD) were compared with a matched group comprising of 84 non ADHD children (Harrier, 2005). Children with ADHD generally prefer extroverted over introverted, practical over imaginative, thinking over feeling, and flexible over organized style s. Although more non ADHD children prefer imaginative over practical style and organized over flexible style, the proportion of children with and without ADHD who displayed preferences for each of the four bipolar traits was comparable. Furthermore, childr en with and without ADHD displayed moderate levels of preferences for all four bipolar traits.

PAGE 33

3 3 Regardless of the diagnosis, gender differences in the preferences for thinking feeling styles were significant. More males displayed a preference for thinking style and more females displayed a preference for a feeling style. In addition, more females displayed a preference for organized styles and more males displayed a preference for flexible styles. Among males, those with ADHD displayed a higher preference for a flexible style than those who did not display ADHD. Thus, while ADHD is not strongly associated with temperament traits in children, the preference among males with ADHD for a flexible style may warrant attention to academic and behavioral accommodat ions in school and home settings (Harrier, 2005) Children with Autism Spectrum Disorder (ASD) The temperament traits of children with high functioning autism were examined, in part, to determine whether a mismatch occurs between temperament styles of A SD children and their parents. Additionally, the study examined whether a mismatch in temperament styles increases parenting stress experienced by parents of children with ASD (Darby, 2009). In general, children with high functioning autism generally prefe r introverted over extroverted, practical over imaginative, feeling over thinking, and organized over flexible styles. Temperament preferences of children with ASD and their parents were more often matched on the extroverted introverted (62%) and organized flexible (53%) styles. On the other hand, the match between temperament preferences of children with ASD children and their parents was less likely on practical imaginative (68%) and thinking feeling (44%) styles. Parents report higher levels of stress wh en they and their children with ASD display similar preferences for thinking feeling styles and lower levels of stress when they and their children display different preferences for thinking feeling styles along with their children with ASD (Darby, 2009).

PAGE 34

34 Children with Learning Disabilities Temperament traits of 84 Canadian children, ages 6 through 14, with diagnosed learning difficulties in reading, spelling, language arts, and arithmetic were compared with a matched group of children without any reporte d learning difficulties (Danielsen, 1991). Children with and without learning difficulties preferred extroverted over introverted, imaginative over practical, feeling over thinking, and flexible over organized styles. However, more children with learning d ifficulties preferred extroverted style while approximately equal proportions of children without learning difficulties prefer extroverted and introverted styles Children Identified as G ifted A number of studies examined the temperament styles of gifted c hildren across gender and ages. In general, gifted children displayed preferences for imaginative over practical (Bireley 1991; Cross, Speirs Neumeister, & Cassidy, 2007; Oakland, Joyce, Horton, & Glutting, 2000) and flexible over organized (Cross, Speirs Neumeister, & Cassidy, 2007; Oakland, Joyce, Horton, & Glutting, 2000) styles. Their preferences for extroverted introverted (Beckner, 1990; Dempsey, 1 975; Gallagner, 1990; Mills, 199 3; Parker & Robinsor, 1989; Oakland, Joyce, Horton, & Glutting, 2000) an d thinking feelings (Beckner, 1990; Gallagner, 1990; Oakland, Joyce, Horton, & Glutting, 2000) styles were somewhat balanced. Gender differences among gifted children generally were apparent in practical imaginative, thinking feeling, and organized flexibl e styles. More females than males prefer imaginative and feeling styles and more males than females prefer thinking and flexible style. Compared to non gifted children, gifted children are more likely to prefer an imaginative style, especially among gifted females (Cross, Speirs Neumeister, & Cassidy, 2007; Oakland, Joyce, Horton, & Glutting, 2000).

PAGE 35

35 Children with Visual I mpairments Temperament traits of visually impaired children ages 10 through 17, were examined across gender and age. In general, children who are visually impaired displayed preferences for practical over imaginative and organized over flexible styles (Oakland, Banner, & Livingston, 2000). Differences in preferences for extroverted introverted styles (Oakland, Banner, & Livingston, 2000; Pi nner & Forlano, 1943; Zahran, 1965) or thinking feeling styles (Oakland, Banner, & Livingston, 2000) were not apparent. Gender differences were apparent. Males were more likely to prefer extroverted, thinking, and flexible styles while females more likely to prefer introverted, feeling, and organized styles. Age differences were apparent in thinking feeling and organized flexible styles. Younger children were more likely to prefer thinking and organized styles while older children more likely to prefer feel ing and flexible styles (Oakland, Banner, & Livingston, 2000). C orrelation between Temperament and Cognitive Variables Te Thus, several studies explored the association of childre cognitive variables. Several studies reported mix results on the relationship between temperament preferences and cognitive variables. One study found consistent positive relationships between introverted style and hievement and intelligence test scores (Myers & McCaulley, 1985). However, another study of elementary and high school students who displayed a preference for extroverted style displayed higher achievement than students who displayed a preference for intro verted style (Tobacyk, Hearn, & Wells, 1990). Children who preferred imaginative and organized styles were found to do better

PAGE 36

36 academically than children who preferred practical and flexible styles ( Fourgurean, Meisgeier, & Swank, 1988; Kaufman, McLean, & L incoln, 1996; Myers & McCaulley, 1985; Tobacyk, Hearn, & Wells, 1990). In addition, students who displayed a preference for thinking styles displayed higher math achievement (Davis, 2007). However, students who displayed a preference for feeling style ha d higher grades than students who display a preference for thinking style. temperament preferences were not related to achievement as measured by California Achievement Test (CTB/MacGr aw Hill, 1985) and to their intelligence as measured by the Wechsler Intelligence Scale for Children Revised (WISC R: Wechsler, 1994). Obtained correlations between extroverted introverted, practical imaginative, thinking feeling, and organized flexible st yles and achievement and IQ generally were low and not significant. Thus, temperament traits may be independent of achievement and intelligence (Oakland, Glutting, & Horton, 1996). In conclusion, some evidences support the relationship between imaginative style, intelligence, and achievement. Studies that reported this significant relationship obtained data from samples with a broad age range (e.g., ages 16 94). Thus, the non significant results obtained by Oakland, Glutting, and Horton (1996) can be expla ined by having younger participants and limited age ra n ge (i.e., ages 8 17). C ross national Differences in Temperament in Children has le d to considerable scholarship, including cross national studies on chi Temperament preferences of children from Australia (Oakland, Faulkner, & Bassett, 2005), Costa Rica (Oakland & Mata, 2007), Gaza (Oakland, Alghorani, & Lee,

PAGE 37

37 2006), Greece (Oakland & Hatzichristou, 2010), Hungary (Katona & Oakla nd, 2000), India (Oakland, S ingh, Callueng, & Goen, 2011 ), Japan (Callueng, de Carvalho, Isobe, & Oakland, 2012), Nigeria (Oakland, Mogaji, & Dempsey, 2006), Pakistan (Oakland, Rizwan, Aftab, & Callueng, 2011), 6), Romania (Oakland, Illiescu, Dinca, & Dempsey, 2009), South Africa (Oakland & Pretorisus, 2009), Samoa (Callueng, Lee Hang, Gonzales, Ling South Korea (Oakland & Lee, 2010), United States (Bassett & Oakland, 2009), Venezuela (Le on Oakland, Wei, & Berrios, 2009), and Zimbabwe (Oakl and, Mpofu, & Sulkowski, 2007) were examined in light of the four bipolar traits measured by the SSQ. A synthesis of data from these 17 countries suggests that children display several prevailing tempe rament qualitie s. For example, children from 13 of the 17 countries generally show a preference for an extroversion style: Australia, Costa Rica, Greece, India, Japan, Pakistan, Samoa, South Kore a, United States, Venezu ela, and Zimbabwe. In contrast, children from Hungary and Nigeria generally show a preference for an introversion style. Children from Gaza and South Africa show a somewhat balanced preference for extroversion and p refe rence for ext roversion increases from 8 to 13 ( Oakland et al., 1996; Bassett, 2005). Gender differences in the thinking and feeling styles are apparent in children fro m Australia, Costa Rica, Gaza, Greece, Hungary, Japan, Repu blic of China, South Africa, Venezuela, and United States. Males generally prefer a thinking style and females generally prefer a feeling style. Both male and female children from Egypt, India, South Africa, South Korea, and Zimbabwe show a general

PAGE 38

38 prefere nce for a feeling style. Samoan male and female children show a preference for a thinking style. Children from 12 countries also show a general preference for a practical style: Gaza Greece, Hungary, India, Japan, Nigeria, Pakistan, hina, Romania, Samoa, South Africa, Venezuela, and Zimbabwe. In contrast, children from Australia, Costa Rica, South Korea, and United States generally show a preference for an imaginative style. Children from Greece have a somewhat balanced preference for practical and imaginative styles. Gender differences are evident among children from China, Pakistan, and United States. In these countries, more males prefer an imaginative style and more females prefer a practical style. Age differences also are evident Increased preference for a practical style is seen in older children from South Korea while increased preference for imaginative style is seen in older children from Egypt and the United States. Children from 16 of the 17 countries generally show a pr eference for an organized style: Australia, Costa Rica, Gaza, Greece, Hungary, India, Japan, Nigeria, Pakistan, Samoa, South Africa, United States ., Venezuela, and, Zimbabwe. In contrast, only children from South Korea show a general preference for a flexible style. Children from Costa Rica, Gaza, Greece, Japan, Nigeria, Pakistan, and the United States show gender related differences in organized flexible styles. Males from these countries are more likely to prefer an or ganized style and females are more likely to prefer a flexible style. Similarly, age related differences are Republic of China, Romania, South Africa, South Korea, Venezuela, a nd Zimbabwe.

PAGE 39

39 Preference of children from these countries for flexible style increases with age. However, a greater proportion of older children from Greece and the United States display a preference for organized style. Test Validity Validity is an essenti degree to which evidence and theory support the interpretation of test scores entailed by proposed uses of test (AERA, APA, NCME, 1999, p. 9). Validity pertains to a judgment about test data obtained for a specific purpose and setting. Test validation is an on going activity of collecting information to increase understanding of test results. The validation process is a shared task of the test developer and the test user. The test developer is responsible in formula ting the conceptual framework and the rationale for a test. On the other hand, a qualified test user conducts research using test data to support the purposes of a test or even expand the uses of an existing test (Urbina, 2004). The following are the sour ces of validity evidence Evidence based on test content Test content includes tasks or questions on a test, including item format and wording. It also covers guidelines and procedures for test administration and scoring. Validity evidence based on test content entails establishing a match or congruence between the content and the construct the test is purported to measure. This approach requires test developers to ensure that test content is adequate and representative of the content domain as well as t he relevance of content domain to test score interpretation. Developing a test specification of the content domain is a common requirement to guarantee adequacy and representative of content domain to what the test intends to measure. Another approach to e stablish validity evidence based on content is through expert judgment. This approach requires experts

PAGE 40

40 to evaluate the extent to which test parts and construct are related (AERA, APA, NCME, 1999). Evidence based on response processes Response processes o f examinees can give information about the congruence of the construct and the response or actual performance of examinees. Evidence based on response processes requires the analysis of responses such as inquiring examinees about their response styles or a ctual responses to items. Information obtained from this procedure can clarify and enhance the definition of the construct. Analyzing patterns of responses in various test parts and how these are related to other individual and situational factors also con tributes to understanding the construct ( AERA, APA, NCME, 1999 ) Group variations in meaning and interpretation of test scores can be clarified by obtaining information response processes. In addition, assessment that requires the recording or evaluating of performance or products of examinees by observers or judges can impact test validity. Hence, it is essential to examine how observers and judges record and evaluate data and how these data are relevant to the test interpretation (AERA, APA, NCME, 1999 ). Evidence based on internal structure Validity evidence that is intended to examine internal structure contributes to the general understanding of the conceptual framework of the test. Evidence of internal structure pertains to whether the correlations am ong test items or test domains fit to the construct on which test interpretation is anchored. The type of analysis used to determine internal structure depends on the purpose of the test. Empirical evidence is provided to support interpretation of test sco res for a multidimensional or a unidimensional internal structure. In some cases,

PAGE 41

41 evidence of internal structure is demonstrated through differential item functioning, That is, items are analyzed to examine if different groups of examinees that have simila r characteristics have different responses on the same item. Differences in item functioning can suggest possible multidimensionality of the internal structure of a test (AERA, APA, NCME, 1999 ). Evidence based on relations to other variables. Correlations of test scores with other variables constitute another source of validity evidence. Evidence based on relations to other variables can be in the form of convergent and discriminant evidence, test criterion relationships, and validity generalization. Conver gent validity refers to the relationships between test scores with other measures that assess similar constructs. Discriminant validity pertains to the relationships of test scores with other measures that assess different constructs. Convergent and discri minant validity evidences can be examined using experimental and correlational research designs (AERA, APA, NCME, 1999 ). Evidence on test criterion relationships has the goal to determine how accurate test scores predict criterion behavior or performance. A criterion variable is a measure of a characteristic or quality that is determined by test users. When conducting a test criterion validity study, findings would be interpreted in light of the relevance and reliability of criterion measure to the intended use of a test. Test criterion relationships are established through concurrent or predictive design. In concurrent validity study, the test scores (predictor) and the criterion variable are obtained at the same time. Predictive validity study assesses the extent to which a test can predict a criterion scores obtained at a later time ( AERA, APA, NCME, 1999)

PAGE 42

42 Validity generalization refers to how accurately findings on test criterion relationships can be applied to a new situation without a systematic invest igation on validity in that new situation. Validity generalization employs meta analytic design that requires collecting validation studies conducted in similar situations and make statistical summaries of these studies that may be useful in calculating te st criterion relationships in a new situation. Meta analysis for the purpose of validity generalization can vary in terms of situational factors: 1) differences in predictor variable measurement, 2) job or curriculum involved, 3) criterion measures used, 4 ) test taker characteristics, and 5) time when the study was conducted. These situational factors are included in meta analysis to determine if variations in these factors have significant effect on test criterion relationships ( AERA, APA, NCME, 1999). Research Objectives and Hypotheses This study examines cross nationally the four factor structure of temperament as measured by the Student Styles Questionnaire. Independent factor solution by country is conducted within the item factor analysis framewor k for categorical or dichotomous variables. Subsequently, measurement invariance is examined in countries that exhibit a satisfactory fit to the four factor model of temperament. In light of the research objectives it is hypothesized that the data from e ach country have a good fit to the four factor model of temperament. It is further hypothesized that the four factor model of temperament is invariant across countries

PAGE 43

43 CHAPTER 2 METHOD S Participants Data were collected on 17,867 children from 21 countrie s as part of an international research program of Dr. Thomas Oakland in collaboration with approximately 50 on site researchers. I also acquired some of the data independently of these efforts. The sample consists of children ages 8 through 17 from diverse geographical regions: Africa (Egypt, Nigeria, and Zimbabwe), Central America (Costa Rica), North America (United States), South America (Brazil and Venezuela), Europe (Hungary, Poland, and Romania), East Asia (China, Japan, and Mongolia), Southeast Asia ( Philippines and Singapore), South Asia (Pakistan), Middle East (Gaza, Iran, and Israel), and Oceania (Australia and Samoa). As seen in Table 2 1, sample sizes of individual countries range from 253 (Israel) to 7,902 (United States). Gender distribution was comparable in almost all countrie s and for the combined group (49.53 % males). The children attended public and private schools and lived mostly in urban settings. Samples from Romania and the U.S. were drawn from their respective standardization samples and stratified according to their national census statistics. Samples from other countries were obtained through research in order to describe the preferred temperament styles of children and to examine possible age and gender differences. Children came f rom families of diverse socio economic status, which was indirectly measured based on type of school (i.e., public and private) and place of residence (i.e., rural and urban). Sample characteristics by country are described below. Australia The Australia sample included 308 children enrolled in public and private primary and secondary schools in the provincial city of Bendigo located in the

PAGE 44

44 number of males and females. Th e sample included children from families who display diverse socio economic status. Brazil The Brazil sample included 4 61 children enrolled in public and private with approximately 55% females. The sample included chil dren from families who display diverse socio economic status. China The China sample included 400 children enrolled in public schools in the through 15 with equal number of males and females. The sample included children from middle to upper lower socio economic families. Costa Rica The Costa Rica sample included 432 children enrolled in public and hrough 15, with approximately 50% males in each age. The sample also included children who generally come from middle class families. Egypt The Egypt sample included 800 children enrolled in public schools in the city of Asyut in the southern region of Up through 15, with approximately 50% males. Children came from middle to upper middle socioeconomic families. Gaza The Palestine sample included 400 children enrolled in public schools in ages ranged from 9 through 17, with equal number of males and females. The sample included children from families who display diverse socio economic status.

PAGE 45

45 Hungary The Hungary sample inc luded 400 children enrolled in public schools. Chi ed from 9 through 16 with equal number of males and females. The sample included children from families who display diverse socio economic status. Iran The Iran sample included 511 children enrolled in public schools. Chi gh 15 with approximately 55% females. The sample included children from families who display diverse socio economic status. Israel The Israel sample included 253 children enrolled in public schools. with approximatel y 50% males. The sample included children from families who display diverse socio economic status. Japan The Japan sample included 493 children enrolled in public schools in Miyakonojo, a city of 170,000 inhabitants and one of the main metropolitan areas of the southern Kyushu region in Japan. approximately 50% males. All children came from lower to upper middle class families. Mongolia The Mongolia sample included 1015 children enrolled in public schools in the cities of Ulaanbaatar and Khovd 16 years, with approximately equal number of ma les and females. The sample included children from lower to upper middle class families. Nigeria The Nigeria sample included 400 chil dren enrolled in public schools Yoruba from the south west region, Ibo from the southeast re gion, and Hausa/Fulani from the north region. The sample included children from families who display diverse socio economic status.

PAGE 46

46 Pakistan The Pakistan sample included 463 children enrolled in private schools in the cities of Karachi, Islamabad, and Que tta. 15, with approximately 50% males for each age group. The sample included children from middle to upper class families. Philippine s The Philippine sample included 400 children enrolled in public schools and priva approximately 50% males. The sample included children from low, middle, and upper class families. Poland The Poland sample included 440 children enrolled in public schools. ges ranged from 9 through 17, with approximately 50% males for each age group. The sample included children from lower to middle class families. Romania The Romania sample included 900 children enrolled in various public es ranged from 9 through 17, with equal number of males and females. Children came from a cross section of socio economic backgrounds. Approximately 86% of the children are of Romanian nationality while the remaining 14% represented other nationalities, in cluding Hungarian and German. Samoa The Samoa sample included 400 children enrolled in a public school in 16, with equal number of males and females. The sample include d children from families who display diverse socio economic status. Singapore The Singapore sample included 483 children enrolled in public with approximately 53% females. The sample included children from middle class families.

PAGE 47

47 United States The United State s sample included 7902 children drawn from the standardization sample of the SSQ and representative of the 1990 U.S. Bureau of the Ce n sus Statistics. Children attending public and private schools locat ed in 29 U.S. states and Puerto Rico age s ranged from 8 through 17 with equal number of males and females. The sample was also representative of major racial/ethnic groups: Whites (n=5,547), African Americans (1,194), and Hispanics (868). The remainin g 2 93 children were from other racial/ethnic groups (e.g., Asian). Venezuela The Venezuela sample included 411 children enrolled in public and private schools located in four major cities: Caracas, Anzoategui, Bolivar, and Zulia. 9 through 16, with approximately equal number of males and females. Children came from middle to upper class families. Zimbabwe Th e Zimbabwe sample included 600 children enrolled in public umber of males and females. Children came from middle to upper class families. Measure The SSQ is a self report paper and pencil group administered measure of temperament traits for students ages 8 through 17. Each of its 69 forced choice items has two alt ernatives that provide for an assessment of preferred behaviors associated with one of four bipolar traits: extroversion introversion (EI), practical imaginative (PM), thinking feeling (TF), and organized flexible (OL). These items were developed and empir ically verified to support the Jungian and Myers Briggs temperament t heory. The EI scale has 23 items, the PM scale has 16 items, the TF scale has 10 items, and the OL scale has 23 items. Additionally, six items provide information simultaneously on two sc ales.

PAGE 48

48 The SSQ can be completed in 20 to 30 minutes Each item requires a student to choose between two options related to the same bipolar dimension (e.g., a thinking option and a feeling option). SSQ scores provide information corresponding to three level s of interpretation. The basic level provides information that focuses on the eight styles and four bipolar traits. The second level provides information pertaining to the combination of two styles. The third level provides information on 16 possible combi nations of style preferences, and is patterned after the Myers Briggs model that combines four styles. The reliability of SSQ scores of U.S. children was established through test retest procedure over a period of seven months using prevalence based standar d scores of 137 students, with ages ranging between 8 through 17. The sample comprised of 75 males and 62 females, with gender distribution approximately equal across age levels. Test retest reliability coefficients were .80 for EI, .67 for PM, .78 for TF, and .74 for OL; with a mean reliability coefficient of .74. Change scores between completions ranged from .50 (EI) through 2.20 (PM) (Oakland, Glutting, & Horton, 1996). The validity of the SSQ was assessed using several methods. Internal validity of the SSQ was established following the substantive construct model of instrument development. Items describing behavioral indicators of the four bipolar traits were developed and subjected to item review by nationally known scholars in temperament theory and l earning styles. A total of 100 items were selected after the initial factor Sequential exploratory factor analysis (EFA) was conducted to address the problem of restricted variance of dichotomous items. First, an EFA was employed directly on the

PAGE 49

49 dichotomous scores with forced rotation of the hypothesized four factors. Results indicated that 69 out of the 100 items met the two criteria set to evaluate factor solution: 0 item loading on the hypothesized factor, and 2) 10% or fewer items cross hypothesized four factor structure. Factor analysis of item parcels was repeated 17 times, with one of t parcels was intended to control the problem of inadequately conditioned correlation matrices as well as confirming if individual items in the opened parcel would load on their hypothesized factor The four factor structure of the SSQ was stable when separate factor analysis using item parcels were conducted across three age groups of students who comprised the standardization sample (Oakland, Glutting, & Horton, 1996) four scales was shown by the generally low intercorrelations, ranging from .03 (for EI and PM) through .24 (for PM and OL). Age, gender, and racial/ethnic differences were examined in several studies to provide s from these studies consistently indicated high agreement on SSQ factor structure and latent means of the four temperament traits across age, gender, and racial/ethnic groups (Strafford & Oakland, 1996 ) Subsequent evaluation of interval validity was con ducted using item parceling to overcome low total variance and item loadings when using dichotomous item scores. Factor loadings in the previous EFA became the basis to form 17 parcels, with four to five items per parcel, distributed across the four SSQ fa ctors: five for EI, four for PM, two for TF, and six for OL. Six items were assigned to two parcels because of their

PAGE 50

50 results, the hypothesized four factor model was supported when using item parcels, of item parcels (Oakland, Glutting, & Horton, 1996). Equivalence at the item level also was assessed through differential item functioning. Response patterns of Hispanic students were similar to those of African American and White students. However, ite ms function somewhat differently on the OL trait between African American and White students (Stafford, 1994) External validation of the SSQ was obtained through contrasted groups, convergent and divergent validation procedures. Using contrasted groups, s temperament preferences were differentiated as a function of their career preferences, classes they liked and disliked, activities they most and least enjoyed, and involvement in special programs in school. Findings indicated that personal (e.g., interest) and contextual (e.g., school programs) qualities significantly influence students temperament style preferences ( Oakland, Stafford, Horton, & Glutting, 2001 ) temperament preferences and values. Correlations between temperament preferences and the values of helpfulness and loyalty were significant. For example, students who display a high preference for thinking style are more likely to value being helpful and students who display a high preference for a feeling style are more likely to value loyalty. Convergent validity also was examined by correlating scores from the SSQ and the Myers Briggs Type Indicator (MBI). Expected results were evident. High correlations

PAGE 51

51 were found on the scales that measure similar qualities while low correlations were found on scales that measure different qualities (Oakland, Glutting, & Horton, 1996) Divergent validity is achieved when a construct measured by a test is independent or not relate d to some hypothesized variables. Evidence of divergent students were not related to their achievement as measured by California Achievement Test (CAT: CTB/MacGraw Hill, 1985) and to their and intelligence as measured by the Wechsler Intelligence Scale for Children Revised (WISC R: Wechsler, 1994). Obtained correlations between scores in the four SSQ traits and achievement and IQ scores generally were low and not significa nt. Thus, temperament constructs measured by the SSQ are independent of achievement and intelligence (Oakland, Glutting, & Horton, 1996) The use of the SSQ in various countries has been increasing. SSQ adaptations have been successfully standardized in R omania, South Korea, and Taiwan. Content validity of the SSQ adapted versions was systematically guided by following the guidelines for test adaptations promulgated by the International Testing Commission (ITC: Hambleton, 2002 ). Factorial structures of th e SSQ scores of children from Australia, China, Costa Rica, Philippines, and Zimbabwe achieved excellent fit while factor structure of SSQ scores from Gaza and Nigeria achieved a modest fit (Benson, Oakland, & Shermis, 2009). Procedure Secondary data are u sed in this study and thus are e xempt from formal review by the University Florida Institutional Review Board (UFIRB). Data gathering in each

PAGE 52

52 country was preceded by evaluating the suitability of the SSQ following the ITC guidelines for test a daptation s R items and direction and determined whether the items were suitable for use in their respective countries. Researchers from Australia, Nigeria, Pakistan, Singapore, and Zimbabwe deemed the SSQ English vers ion appropriate for use because target participants in these countries are fluent in the E nglish language and have at least a third grade Englis h reading level. Researchers in countries where English was n ot the primary language formed advisory group s con sisting of bilingual psychologists and educators who are fluent in both their native language and English to translate the SSQ following the back translation sequential method. The procedure essentially involves one or two bilingual speakers to translate t he SSQ to the native language of the children. Then different bilingual speakers back translated the SSQ into English. Discrepancies between the original English version and translated version of the SSQ were discussed and a consensus agreement made on th e most developmentally and cultur ally appropriate translation. Countries in which the SSQ was translated into native languages include Brazil, China, Costa Rica, Egypt, Gaza, Hungary, Iran, Israel, Japan, Mongolia, Philippines, Poland, Romania, Samoa, and Venezuela Table 2 2 reports the translation languages used and on site researchers from 21 countries included in the study. Participation of children was voluntary. Informed consent was obtained from the parents or guardians of the participating children by the researchers. Data gathering and control for extraneous testing factors were facilitated by request from the researchers to school heads to allow them to administer the SSQ to children in their

PAGE 53

53 respective classrooms. Children indicated their answers either using a separate response form or directly on the item booklet. Test administration was conducted either by the researchers or trained research assistants who followed administration procedures stated in the SSQ manual. Data Analyses Different stat istical procedures were used to address the goals of t he study. These procedures utilize d either item scores or trait scores of the SSQ. The six SSQ items assigned to two bipolar traits were not included in the analysis to control the problem of cross item variance. Thus, data from 63 items with unique factor solutions were included in the analysis. Data Screening Missing da ta in each country data set was handled in the following manner. First children who did not answer six or more items from the SSQ wer e dropped from further analysi s. Second, when calculating trait scores, children who missed to answer more than one item in any temperament trait will be treated as missing data in that sp ecific trait (Schimitt, Allik, McCrae, & Benet Martinez, 2007). Chi ldren with missing data will not be included in the analysis. Internal Consistency of the SSQ Preliminary analyses included examining internal consistency using Cronbach alpha and tetrachoric correlation of SSQ scores from eac h country. Cronbach alphas w ere calculated for each of the four temperament traits using dichotomous item scores and following the meth od used by Fan & Thomson (2001). A 95% confidence interval was established for each alpha coefficient to account for the band of values that accurat ely estimate s internal consistency of dichotomous item scores.

PAGE 54

54 Structural Validity Structural equivalence of the hypothesized four factor structure was examined in U.S. children across age, gender, and racial/ethnic groups using EFA (Stafford, 1994 ; Staff ord & Oakland, 1998 ) and across countries using CFA ( Benson, Oakland, & Shermis, 2009 ). Both studies used item parcels instead of individual items as observed or manifest variables to remedy the problem of restricted variance of dichotomous items. Each par cel contained four to five items that m easure the same trait. In forming the parcels, items with high and low p values were grouped together to counterbalance the d ifficulty levels within each parcel (Oakland, Glutting, & Horton, 1996). Item parceling has been recommended as a method for lengthy tests with categorical or ordinal items when subjected to structural equation modeling (Yang, Nay, & Hoyle, 2010). Use of ite m parcels in factor analysis also has empirical advantages that include increasing reliab ility, meeting normality assumptions, resolving concerns on small sample sizes, simplifying interpretations, and achieving better model fit (Bandalos & Finney, 2001). Item parceling is not without criticisms. Empirical arguments for the use of item parce ls in psychometric analysis are associated with the problem of dimensionality of the construct, the misspecification of factor model, and the inaccurate meaning of parameter estimates (Little, Cunningham, Shahar, & Widaman, 2002). Parceling is problematic because items in any parcels may represent different meanings other than the hypothesized construct. Because of the multidimensionality of the item parcels, measurement models may have biased loading estimates that may have serious implications in explaini ng the variance of the latent construct (Badalos & Finney, 2001). Misspecification of factor model occurs when there is no information of the manifest

PAGE 55

55 variable (i.e., item) that underlie a latent construct. Hence, the measurement model is misspecified on a ccount of the variance explained by individual items that directly describe the construct being measured. The factor structure of the SSQ was highly similar across age, gender, and racial/ethnic groups (Stafford, 1994; Stafford & Oakland, 1998). Likewise independent CFA of the SSQ data from seven countries replicated the four factor structure of the SSQ. However, partial invariance was achieved when the data were examined for multigroup CFA (Benson, Oakland, & Shermis, 2009) In the current study, the c ategorical nature of the SSQ item options was considered when determining the factor structure of temperament traits cross nationally. Using Mplus 6.1 (Muthn & Muthn, 2006), a tetrachoric correlation matrix of item data for each country was fitted for CF A using a mean and variance adjusted weighted least squares estimator (WLSMV) with theta parameterization. The WLSMV is an advanced estimator for non normal data when multivariate normality is not assumed. In addition, the WLSMV is a highly efficient estim ator for non normal data compared to other asymptotically distribution free (ADF) estimators (Muthn & Muthn, 2006). When using WLSMV, the root mean square error approximation (RMSEA) is it ( Loehlin, 1998; Yu & Muthen, 2002) Comparative fit index (CFI) and Tucker Lewis Index (TLI), commonly used model fit indexes, do not perform well in WLSMV and thus, were considered secondary criteria (Ivanova et al., 2007), in this study. Hu and Bentle r (1999) suggested er, alternative criteria

PAGE 56

56 90 for a good fit and .80 to .89 for an acceptable fit were adopted for use in this study (Brown & Cudeck, 1993 ; Marsh & Hau, 1996 ). S eque ntial multigroup CFA was implemented starting with a single group analysis in each country followed by a multiple group analysis A CFA for each country was conducted to examine the factor structure of temperament as measured by the SSQ. Only the countries were included in the multigro up CFA. Multigroup analysis was conducted to assess factorial invariance between a target country (e.g., Australia) and a reference country (U.S.). The U.S. is chose n as the reference country because the SSQ was developed in the U.S. and a stable SSQ four factor structure has been established using data from more than 7,000 U.S. children. The test of factorial invariance is a two group CFA (i.e., a target country an d the U.S. as the refere nce country) that examine s a four factor structure of temperament in terms of the configural, metric, and scalar invariance (Dimitrov, 2010). A configural invariance is tested by fitting a baseline model with the same patterns of ze ro and non zero loadings for a target country and the U.S. A metric invariance is tested by constraining the factor loadings to be equal between a target country and the U.S. A scalar invariance is tested by constraining both the factor loadings and the it em intercepts to be equal for a target country and the U.S. (Meredith, 1993; Muthn & Muthn, 2006). Metric and scalar invariance models are tested at the same step because when using categorical data, factors loadings and item intercepts must be constrain ed in tandem (Muthn & Muthn, 2006).

PAGE 57

57 The multigroup CFA was conducted following important assumptions. First, each item would be associated with only the factor it was intended to measure. Second, the covariation of the four factors of temperament would b e allowed. Third, the post hoc model fitting would be set to a minimum and correlated error terms would be allowed when supported by strong empirical evidence. Fourth, the group specific error covariance would be left unconstrained for a target country and the U.S. through the factorial invariance testing procedure (Byrne, 1994; Byrne, Shavelson, & Muthen, 1989). Factorial invariance is determined through the statistical significance of the difference in the chi square ( 2 ) values and CFI between the two n ested mode ls. A partial invariance is conducted if the difference test result is significant based on the modification indices (MI > 10) information ( Cheung & Rensvold, 2002; Vandenberg & Lance, 2000).

PAGE 58

58 Table 2 1. Sample sizes of countries by total, gender and age group Country Total Gender Age Group Male Female 8 10 11 12 13 14 15 17 Australia 308 157 151 78 78 74 78 Brazil 461 212 249 124 139 114 84 China 400 200 200 100 100 100 100 Costa Rica 432 212 220 108 108 108 108 Egypt 800 399 401 178 197 213 212 Gaza 400 200 200 100 100 100 100 Hungary 401 201 200 99 100 102 100 Iran 511 237 274 133 161 140 77 Israel 253 114 139 87 166 Japan 493 223 270 149 141 93 112 Mongolia 1009 502 507 251 245 252 261 Nigeria 400 200 200 100 100 100 100 Pakistan 463 232 231 126 118 111 108 Philippines 400 200 200 100 100 100 100 Poland 440 202 238 44 155 73 168 Romania 900 450 450 225 225 225 225 Samoa 400 200 200 100 100 100 100 Singapore 483 255 228 193 139 90 61 United States 7,902 3,950 3,952 2 ,111 2,293 2,083 1,415 Venezuela 411 203 208 101 100 102 108 Zimbabwe 600 300 300 150 150 150 150 Total 17,867 8,849 9,018 4,657 5,015 4,430 3,767

PAGE 59

59 Table 2 2 SSQ translation languages, names and institutional affiliations of on site researchers Coun try Translation language Lead on site researcher Institutional Affiliation Australia Michael Faulkner, Ph.D. La Trobe University Brazil Portuguese Raquel Guzzo, Ph.D. Pontifucal Catholic University of Campinas China Mandarin Li Lu, Ph.D Shanxi Medical University Costa Rica Spanish Ana Mata, M.A. Egypt Arabic Mahmoud Emam, Ph.D. Assiut University Gaza Arabic Mohammed Adnan Alghorani, Ph.D. United Arab Emirates University Hungary Hungarian Nora Katona Eotvos Lorand University Iran Arabic Mohammed Adnan Alghorani, Ph.D. United Arab Emirates University Israel Hebrew Sharone Maital, Ph.D. University of Haifa Japan Nihonggo Moiss Kirk de Carvalho Filho, Ph.D. Kyoto University Mongolia Mongolian Seded Bathkuyag, Ph.D. National University of Mongolia Nigeria Andrew Mogaji, Ph.D. University of Lagos Pakistan Muhammad Rizwan, Ph.D. University of Karachi Philippines Tagalog Carmelo Callueng, M.S. De La Salle University Poland Polish Tomasz Rowinski, Ph.D. Cardinal Ste fan Wyszynski University Romania Romanian Dragos Iliescu, Ph.D. SNSPA/D & D Research Samoa Samoan Desmond Lee Hang, Ph.D. National University of Samoa Singapore Yoke Fong Lau M.Ed University of Florida Venezuela Spanish Carmen Leon, Ph.D. Uni versidad Catolica Andres Bello Zimbabwe English Elias Mpofu, Ph.D. Pennsylvania State University Note: Countries with dash ( ) used English language version of the SSQ

PAGE 60

60 CHAPTER 3 RESULTS Data Screening Before subjecting the data to CFA, data scree ning was conducted for each country data set to identify cases with at least 10% (6 items) of missing responses. Of the 21 countries included in the study, nine (43%) did not have a single case with substantial missing responses. The remaining 12 (57%) cou ntries have missing data items) was very minimal. Table 3 1 reports the number of cases per country with missing data. Internal Consistency As re ported in Table 3 2 Cr onbach alpha was calculated as a measure of internal consistency of item respo nses for each temperament trait per country. Using data in 16 countries display acceptable consistency. TF data in 2 countries display acceptable level of internal consistency. OL data in 16 countries dis play acceptable levels of internal consistency. Data from Egypt, Mongolia, and Samoa display unacceptable estimates of internal consistency across the four temperament factors. Structural Validity The fit of the four factor model of temperament as measure d by the SSQ was examined through the use of confirmatory factor analysis (CFA) for each country. Tetrachoric correlation matrix was fitted to the model, and the parameters were

PAGE 61

61 estimated using the mean and variance adjusted weighted least squares (WLSMV) estimator with theta parameterization. An initial CFA solution used the default paths of Mplus 6.1 with the first factor loading of each set of parameters automatically constrained to 1.0. However, using this default version resulted to non convergence sol ution for data sets from Brazil, China, Costa Rica, Philippines, and Samoa. As an alternative, the variance of the latent factors was fixed to 1.0 for all country data sets. As mentioned in the previous chapter, the overall goodness of fit was evaluated us ing multiple crite ria: 1) a non significant WLSMV 2 fit index (Loehlin, 1998; Yu & Muthen, 2002) secondary model fit indices (Brown & Cudeck, 1993 ; Marsh & Hau, 1996 ). As reported in Table 3 3 CFA results were consistent f or all the countries. That is, 2 values were significant, RMSEA values were less than .06 and both the CFI and TLI values were less than .90. RMSEA values indicate that each country data set has a good fit to th e four factor model of temperament. However, both the CFI and TLI indicate that the model did not fit the data for each of the 21 countries. The statistically significant 2 values are due to large sample size. Most item indicat ors converged on the traits they intended to measure. However, factor loadings generally were low. In most countries, the bipolar traits are distinct on what they measure. Taken together, the results of the model fit indices indicate that the four factor model of temperament did not fit the data reasonably well for each of the 21 countries. Hence, the hypothesis that the four factor model of temperament is not confirmed. The lack of model fit precludes tests of cross national invariance of the four factor model of temperament.

PAGE 62

62 CFA fo r China, Egypt, and Gaza reported a non positive definite matrix that was likely the result of an observed linear dependency of factors. As a result, a series of exploratory factor analysis (EFA) was performed for these three countries to determine an alte rnative factor solution. Some items have significant loadings on factors other than what they intend to measure. These misspecified items were more frequent in EI (e.g., items 40 and 59) and OL (e.g., items 1 and 62). To address item misspecifications and consequently improve model fit, a modified four factor solution was performed for each of the 18 countries with positive definite matrix This was done by allowing some items to cross load on other factors. A modified solution was not done with the Samoa data because the misspecified items were not logically and substantively related to other factors that they seem to load. The modified solution did not significantly improve model fit in all the 18 countries. Similar to the initial CFA solution, the WLSMV 2 values were statistically significant RMSEA values were within acceptable range indicating good model fit, and the CFI and TLI were consistently below .90 indicating non fit to the model A report on the structural validity of the SSQ is presented bel ow for each country. Included in the report is a summary of factor loadings, standard error, and threshold of item indicators. Australia (N = 368) Australian children prefer extroverted more than introverted, imaginative more than practical, thinking mor e than feeling and flexible more than organized styles. The initial CFA solution yielded a significant 2 (1884) = 2606.560, p < .001; RMSEA = .032; CFI = .733; and TLI =.723. As a whole, these fit statistics suggest that the data from Australian children did not exhibit a good fit to the four factor structure of

PAGE 63

63 temperament. An inspection of the squared sta ndardized factor loadings of item indicators revealed that 53 (17 EI, 9 PM, 7 TF, and 20 OL items) out of 63 item indicators have appreciable loadings (i.e., ) to their associated factors. Each of these items contributed a significant amount of varian ce that can be explained by the factor it is inten ded to measure. The remaining 10 items (2 EI, 5 PM, 1 TF, and 2 OL items) did not have sub stantial factor loadings (i.e., < .30 ) to their associated factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor. Factor correlations range from low (r = .074 for EI and TF) to moderate (r = .508 for PM and OL), with most correlations low yet signifi cant. Factor correlations for are related (e.g., PM and OL), they are not overlapping and thus are distinct in terms of the construct they measure. A review of t he initial CFA solution indicated significant misspecifications of factorial structure for nine items: 1, 11, 21, 28, 40, 51, 52, 55, and 62. Allowing some of these items to cross load on other factors would likely improve the model fit statistics. After e valuating the content of these items, three items appeared to have substantive and logical meaning to factors other than what these items intend to measure. Consequently, items 28 and 40 were allowed to cross load on EI factor and item 1 was allowed to cro ss load on OL factor. The resulting model fit statistics of the modified CFA solution for the Australian data yielded a significant 2 (1881) = 2548.690, p < .001; RMSEA = .031; CFI = .753; and TLI =.744. These fit statistics suggest that the data from Aust ralian children did not exhibit a good fit to the four factor structure of

PAGE 64

64 temperament, al beit with modifications. Table 3 4 reports the par ameter estimates of the four factor solution in the Australia data. Brazil (N = 461) Brazilian children prefer ext roverted more than introverted, imaginative more than practical, and organized more than flexible styles. Approximately equal proportion of Brazilian children prefers thinking and feeling styles. The initial CFA solution yielded a significant 2 (1884) = 2990.948, p < .001; RMSEA = .033; CFI = .639; and TLI =.626. As a whole, these fit statistics suggest that the data from Brazilian children did not exhibit a good fit to the four factor structure of temperamen t. An inspection of the squared stan dardized factor loadings of item indicators revealed that 50 items (18 EI, 10 PM, 8 TF, and 14 OL items) have appreciable loadings (i.e., .30 ) to their associated factors. Each of these items contributed a significant amount of variance that can be expla ined by the factor it is intended to measures. The remaining 13 items (1 EI, 4 PM, and 8 OL items) did not have substantial factor loadings (i.e., < .30 ) to their associated factors and conversely have very high residual variances. Each of these items acco unted for a significant proportion of variance that is not explained by its associated latent factor. Factor correlations generally are low, ranging from r = .059 for TF and OL to r = .252 for PM and OL. This suggests that factor correlations for the B razil data indicate the constructs they measure. A review of the initial CFA solution indicated significant misspecifications of factorial structure for items 1, 15, 16, 23, 33, 39, 47, 48, 50, 57, 59, 62, 64, 65, and 66. Allowing some of these items to cross load on other factors would likely improve the

PAGE 65

65 model fit statistics. After evaluating the content of these items, three items appeared to have substantive and logical meaning to factors other than what these items intend to measure. Consequently, item 59 was allowed to cross load on EI factor and items 1 and 21 were allowed to cross load on OL factor. The resulting model fit statistics of the modified CFA so lution yielded a significant 2 (1881) = 2893.080, p < .001; RMSEA = .033; CFI = .67; and TLI =.658. These fit statistics suggest that the data from Brazilian children did not exhibit a good fit to the four factor structure of temperament, albeit with modif ications. Table 3 5 reports the par ameter estimates of the four factor solution for the Brazil data. China (N = 399 ) Chinese children prefer practical more than imaginative, thinking more than feeling, and organized more than flexible styles. Approximate ly equal proportion of Chinese children prefers extroverted and introverted styles. The initial CFA solution yielded a significant 2 (1884) = 2754.924, p < .001; RMSEA = .034; CFI = .741; and TLI =.731. As a whole, these fit statistics suggest that the data from Chinese children did not exhibit a good fit to the four factor structure of temperament. Moreover, the initial solution indi cates a non positive definite correlation matrix that may be due to very high correlation coefficients that indicate substantial overlap of some factors. Specifically, correlation coefficients are 1.060 for PM and TF and 1.050 for TF and OL. All other int er factor correlation coefficients range from r = .233 (EI and OL) to r = .789 (PM and OL). Although sample Pearson product moment correlation coefficients cannot be more extreme than 1.0 in CFA sample inter factor correlation coefficients can be more e xtreme than 1.0 Such estimates are called inadmissible or improper and indicate a potential problem with the model even when

PAGE 66

66 goodness of fit indices are adequate. Because the CFA resulted to an improper solution, an EFA was conducted to determine an alter native model that may be salient and meaningful to the data in Chinese children. Among the EFA solutions for two to five factors, the four factor solution seemed to provide a more logical and meaningful convergence of items de spite a poor fit to the model : 2 (1767) = 2288.786, p < .001; RMSEA = .027; CFI = .845; and TLI =.828. Factor 1 has 23 items with subst antial factor loadings (i.e., .30 ), with 18 items (78%) measuring OL trait and the remaining 5 items (28%) measuring EI, PM, or TF trait. Factor 2 has 17 items with subst antial factor loadings (i.e., .30 ), with 10 items (59%) measuring PM trait and the remaining 7 items (41%) measuring EI, TF, or OL trait. Factor 3 has 18 items with subst antial factor loadings (i.e., .30 ) with 16 items (89%) mea suring EI trait and the remaining 2 items (11%) measuring TF or OL trait. Seven items (i.e., 17, 23, 25, 32, 62, 65, & 66) loaded on more than one factor. Inter factor correlation are low ranging from r = .072 to r = .187. In summary, the three factor so lution represents a stable structure for items that measure OL, PM, and EI traits. Items that measure TF trait did not cluster together to form a stable and independent factor. Table 3 6 reports the EFA model fit indices and number of items with significan t loadings for the different factor solutions in the China data. Costa Rica (N = 431) Costa Rica children prefer extroverted more than introverted, practical more than imaginative, feeling more than thinking, and organized more than flexible styles. The initial CFA solution yielded a significant 2 (1884) = 3164.511, p < .001; RMSEA = .026; CFI = .590; and TLI =.575. As a whole, these fit statistics suggest that

PAGE 67

67 the data from Costa Rica children did not exhibit a good fit to the four factor structure of temperamen t. An inspection of the squared sta ndardized factor loadings of item indic ators revealed that 39 items (14 EI, 6 PM, 6 TF, and 13 OL items) have app reciable loadings (i.e., .30 ) with their associated factors. Each of these items contributed a significant amount of variance that can be exp lained by the factor it is intended to me asure. The remaining 24 items (8 EI, 8 PM, 2 TF and 9 OL items) did not have subst antial fac tor loadings (i.e., < .30 ) to their intended factors and conversely displayed with very high residual variances. Each of t hese items accounted for a significant proportion of variance that is not explained by its associated factor. Most factor correlations are significant and range from very low (r = .077 for PM and TF) to moderate (r = .439 for PM and OL). This suggests that factor correlations for factors are distinct on the constructs they measure. A review of the initial CFA solution indicated significant misspecifications of factor ial structure for several items. Allowing some of these items to cross load on other latent factors would likely improve the model fit statistics. After evaluating the content of these items, four items seemed to have substantive and logical meaning to oth er later factors. Consequently, items 18 and 23 were allowed to cross load on EI factor and items 36 and 62 were allowed to cross load on OL factor. The resulting model fit statistics of this modified CFA solution yielded a significant 2 (1880) = 3088.953, p < .001; RMSEA = .025; CFI = .613; and TLI =.598. These fit statistics suggest that the data from Costa Rica children did not exhibit a good fit to the four factor structure of

PAGE 68

68 temperament, albeit with modificati ons. Table 3 7 reports the par ameter estimates of the four factor solution in the Costa Rica data. Egypt (N = 920) Egyptian children prefer extroverted more than introverted, practical more than imaginative, feeling more than thinking, and organiz ed more than flexible styles. The initial CFA solution yielded a significant 2 (1884) = 3146.025, p < .001; RMSEA = .027; CFI = .513; and TLI =.495. As a whole, these fit statistics suggest that the data from Egyptian children did not exhibit a good fit to the four factor structure of temperament. Moreover, the initial solution ind icates a non positive definite correlation matrix that may be due to a very high correlation coefficient for TF and OL (r = 1.051). Also, PM highly correlates with TF (r = .859) and OL (r = .795). These high correlation coefficients indicate substantial overlap on what PM, TF, and OL intended to measure. All other inter factor correlation coefficients are moderate and range from r = .393 (EI and TF) to r = .556 (EI and OL). Although sample Pearson product moment correlation coefficients cannot be more ex treme than 1.0 in CFA sample inter factor correlation coefficients can be more extreme than 1.0 Such estimates are called inadmissible or improper and indicate a potential problem with the model even when goodness of fit indices are adequate. Because t he CFA resulted in an improper solution, an EFA was conducted to determine an alternative model that is salient and meaningful to the data. Among the EFA solutions for two to five factors, no factor solution showed a discernible pattern of item clustering and model fit indices for all factor solutions generally are not adequate. In the two factor solution ( 2 (1828) = 2795.259, p < .001; RMSEA = .024; CFI = .627; and TLI =.601), 16 items hav e substantial loadings (

PAGE 69

69 .30 ) on factor 1 with 10 or 63% o f the items measuring OL trait and four or 25% of the items measuring PM trait. The remaining two items measure EI and TF traits. For factor 2, 10 items have substantial loadings, with half of the items measuring OL trait and the other items measuring EI, PM, and TF traits. Item 15 cross loaded on factors 1 and 2. Using a three factor solution ( 2 (1767) = 2496.157, p < .001; RMSEA = .021; CFI = .718; and TLI =.689), 18 items hav e substantial loadings (i.e., .30 ) on factor 1, with 11 or 61% of the items measuring OL, five or 28% of the items measuring PM, one item each measuring EI and TF. For factor 2, 10 items have substantial loadings, with half of the items measuring OL and the other items measuring EI, PM, or TF traits. For factor 3, all nine items that have substantial factor loadings measure EI trait. Two items (i.e., item # 42 and 57) c ross load on factors 1 and 3. Using a four factor solution ( 2 (1707) = 2235.302, p < .001; RMSEA = .018; CFI = .796; and TLI =.767), 10 items hav e substantial loadings (i.e., .30 ) on factor 1, with more than half (60%) of the items measuring OL and the other items measuring PM or TF. For factor 2, 11 items have substantial loadings mostly measuring PM and OL traits. For factor 3, nine items have substantial factor loadings that measure EI, PM, TF, or OL traits. For factor 4, nine of the 10 items that have substantial factor loadings measure EI trait. Four items (i.e., item # 15, 32, 47 and 66) cross load on more than one factor. Using a five factor solution ( 2 (1648) = 2062.220, p < .001; RMSEA = .017; CFI = .840; and TLI =.810), 16 items hav e substantial loadings (i.e., .30 ) on factor 1, with half of the items measuring OL and the other items measuring EI, PM, or TF. For factor 2, three items have substant ial loadings that measure PM or TF. For factor 3, 11 items have substantial factor loadings, with most items measuring OL and few items

PAGE 70

70 measuring EI, PM, or TF. For factor 4, six of the eight items with substantial factor loadings measure OL and the remai ning two items measure PM. For factor 5, all eight items with substantial factor loadings measure EI. Nine items (i.e., item # 3, 15, 20, 31, 32, 42, 47, 57, and 66) cross load on more than one factor. Table 3 8 reports the EFA model fit indices and numbe r of items with significant loadings for the different factor solutions in the Egypt data. Gaza (N = 400) Gaza children prefer extroverted more than introverted, practical more than imaginative, thinking more than feeling, and organized more than flexibl e styles. The initial CFA solution yielded a significant 2 (1884) = 2461.500, p < .001; RMSEA = .029; CFI = .755; and TLI =.746. As a whole, these fit statistics suggest that the data from Gaza children did not exhibit a good fit to the four factor structure of temperament. Moreover, the initial solution indicat es a non positive definite correlation matrix that may be due to very correlation coefficients and indicate substantial overlap of some factors. Most inter factor correlation coefficients are evident of multicollinearity ghly correlates with EI (r = .904), PM (r = .890), and OL (r = .938). OL also highly correlates with EI (r = .852) and PM (r = .806). Correlation between EI and PM (r = .605) is moderate. A very high correlation of factors in CFA indicates a potential problem with the model even when goodness of fit indices are adequate. Because the CFA resulted to an improper solution, an EFA was conducted to determine an alternative model that is salient and meaningful to the data in Gaza children.

PAGE 71

71 EFA solutions for two to five factors were run and among these solutions, the two factor solution seemed to provide a more logical and meaningful convergence of items despite a poor fit to the model: 2 (1828) = 2152, p < .001; RMSEA = .022; CFI = .862; and TLI =.853. Facto r 1 has 33 items with subst antial factor loadings (i.e., .30 ) that mostly measure OL (19 items or 58%) and PM (8 items or 24%). The remaining six items are equally divided to measure EI and TF. In factor 2, 13 out of 14 items have subst antial f actor load ings (i.e., .30 ) that measure EI and one item measuring PM. The correlation between factors 1 and 2 is low (r = .090). In summary, the two factor solution represents a stable structure for items that measure EI on one factor and a combination of items that measure PM and OL on the other factor. Very few TF items have significant loadings and did not cluster as an independent factor. Table 3 9 reports the EFA model fit indices and number of items with significant loadings for the different factor soluti ons in the Gaza data. Hungary (N = 494) Hungarian children prefer extroverted more than introverted and organized more than flexible styles. Approximately equal proportions of Hungarian children prefer practical and imaginative as well as thinking and f eeling styles. The initial CFA solution yielded a significant 2 (1884) = 2617.849, p < .001; RMSEA = .031; CFI = .475; and TLI =.456. As a whole, these fit statistics suggest that the data from Hungarian children did not exhibit a good fit to the four fact or structure of temperamen t. An inspection of the squared standardized factor loadings of item indicators revealed that 37 items (15 EI, 5 PM, 4 TF, and 12 OL items) have a ppreciable loadings (i.e., .30 ) to their associated factors. Each of these items c ontributed a significant amount of variance that can be explained by the factor it is inten ded to

PAGE 72

72 measure. The remaining 26 items (4 EI, 9 PM, 4 TF and 10 OL items) did not have subst antial factor loadings (i.e., < .30 ) to their intended latent factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated latent factor. Factor correlations for EI and PM (r = .049), EI and TF (r = .086), and EI and OL (r = .157) are in the low range. Moreover, significant and moderate correlations are .85) is not evident on these correlations. EI and TF do not overlap with PM and OL and are distinct in terms of the constructs they measure. On the other hand, the correlation between PM and OL (r = .867) provides evidence of multicollinearity and suggests an overlap on what PM and OL measure. A review of the initial CFA solution indi cated significant misspecifications of factorial structure for items 22, 38, 62, and 67. Allowing some of these items to cross load in other latent factors may likely improve the model fit statistics. After evaluating the content of these items, three item s seemed to have substantive and logical meaning to other later factors. Consequently, items 62 and 67 were allowed to cross load on OL and TF factors respectively. The resulting model fit statistics of this modified CFA solution for the Hung ary data yield ed a significant 2 (1882) = 2592.397, p < .001; RMSEA = .031; CFI = .492; and TLI =.472. These fit statistics suggest that the data from Hungarian children did not exhibit a good fit to the four factor structure of temperament, alb eit with modifications. Table 3 10 reports the par ameter estimates of the four factor solution in the Hungary data.

PAGE 73

73 Iran (N = 510) Iranian children prefer practical more than imaginative and organized more than flexible styles. Approximately equal proportions of Iranian children prefer extrovert ed and introverted as well as thinking and feeling styles. The initial CFA solution yielded a significant 2 (1884) = 2878.989 p < .001; RMSEA = .032; CFI = .722; and TLI =.712. As a whole, these fit statistics suggest that the data from Iranian children did not exhibit a good fit to the four factor structure of temperamen t. An inspection of the squared stand ardized factor loadings of item indicators revealed that 36 items (8 EI, 7 PM, 2 TF, and 19 OL items) have appreciable loadings (i.e., ) to their intended factors. Each of these items contributed a significant amount of variance that can be explained by the factor it is intended to measure. The remaining 27 items (11 EI, 7 PM, 6 TF, and 3 OL items) did not have subst antial factor loadings (i.e., < .30 ) to their intended factors and conversely have very high residual variances. Each of these items acco unted for a significant proportion of variance that is not explained by its associated latent factor. Factor correlations range from moderate (r = .508 for PM and TF) to high (r = .879 for EI and PM). The high correlation between EI and PM indicates mul ticollinearity are in the moderate range and the correlations are distinct in the construct they measure. A review of the initial CFA solution indicated a significa nt misspecification of factorial structure for item 27. When item 27 was allowed to cross load on PM factor, the resulting model fit statistics of the modified CFA solution yielded a sign ificant 2 (1883) = 2866.009, p < .001; RMSEA = .032; CFI = .725; and TLI =.715. These fit statistics

PAGE 74

74 suggest that the data from Iranian children did not exhibit a good fit to the four factor structure of temperament, al beit with modification. Table 3 11 reports the par ameter estimates of the four factor solution in the Iran data. Israel (N = 218) Israeli children prefer extroverted more than introverted, practical more than imaginative, feeling more than thinking, and organized more than flexible styles. The initial CFA solution yielded a significant 2 (1884) = 2175.017 p < .001; RMSEA = .027; CFI = .749; and TLI =.739. As a whole, these fit statistics suggest that the data from Israeli children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the squared stand ardized factor loadings of item indicators revealed that 44 items (15 EI, 8 PM, 7 TF, and 14 OL items) h ave appreciable loadings (i.e., ) on their intended factors. Each of these items contributed a significant amount of variance that can be explained by the factor it is inten ded to measure. The remaining 19 items (4 EI, 6 PM, 1 TF, and 8 OL) did not have significant factor loadings (i.e., < .30 ) on their intended factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor. Moderate and significant correlations are reported for PM and TF (r = .386) and for PM and OL (r = .432). All other factor correlations are low and not significant. Mult are distinct and do not overlap in the construct they measure. A review of the initial CFA solution indicated significant misspecifications of factorial structur e for items 1, 50, and 68. Allowing these items to cross load on other

PAGE 75

75 latent factors would improve the model fit statistics. However, only items 1 and 50 seemed to have substantive and logical meaning associated with OL and EI, respectively. When item 1 w as allowed to cross load on OL trait and item 50 was allowed to cross load on EI trait the resulting model fit statistics of the modified CFA solution for the Israel data yielded a significant 2 (1882) = 2144.574, p < .001; RMSEA = .025; CFI = .773; and TLI =.763. These fit statistics suggest that the data from Israeli children did not exhibit a good fit to the four factor structure of temperament, albeit with modifications. Table 3 12 reports t he par ameter estimates of the four factor solution in the Israel data. Japan (N = 494) Japanese children prefer practical more than imaginative and organized more than flexible styles. Approximately equal proportions of Japanese children prefer extrovert ed and introverted as well as feeling and thinking styles. The initial CFA solution yielded a significant 2 (1884) = 2904.025 p < .001; RMSEA = .033; CFI = .732; and TLI =.722. As a whole, these fit statistics suggest that the data from Japanese children did not exhibit a good fit to the four factor structure of temperamen t. An inspection of the squared stan dardized factor loadings of item indicators revealed that 50 items (18 EI, 8 PM, 7 TF, and 17 OL items) h ave appreciable loadings (i.e., ) to their intended factors. Each of these items contributed a significant amount of variance that can be explaine d by the factor it is intended to measure. The remaining 13 items (1 EI, 6 PM, 1 TF, and 5 OL items) did not have signi ficant factor loadings (i.e., < .30 ) to their intended factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor.

PAGE 76

76 Factor correlations are generally low and range from r = .028 (EI and OL) to r = .199 (TF and OL), with most correlations not significant. This implies that facto r distinct in terms of the construct they measure. A review of the initial CFA solution indicated significant misspecifications of factorial structure for items 9, 11, 18, 22, 23, 28, 31, 34, 35, 40, 43, 45, 50, 55, 61, 62, 63, and 68. Allowing some of these items to cross load on other latent factors may improve the model fit statistics. After evaluating the content of these items, items 31, 50, and 63 appeared t o have substantive and logical meaning to other factors other than the factors they are associated to measure. Consequently, item 31, 50, and 63 were allowed to cross load on OL, EI, and PM, respectively. The resulting model fit statistics of the modified CFA solution for the Ja pan data yielded a significant 2 (1881) = 2861.529, p < .001; RMSEA = .033; CFI = .743; and TLI =.733. These fit statistics suggest that the data from Japanese children did not exhibit a good fit to the four factor structure of tempe rament, alb eit with modifications. Table 3 13 reports the par ameter estimates of the four factor solution in the Japan data Mongolia (N = 1,003) Mongolian children prefer practical more than imaginative, thinking more than feeling, and organized more than flexible styles. Approximately equal proportion of Mongolian children extroverted and introverted styles. The initial CFA solution yielded a significant 2 (1884) = 3166.944, p < .001; RMSEA = .026; CFI = .589; and TLI =.574. As a whole, these fit statistics suggest that the data from Mongolian children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized factor loadings of item indicators

PAGE 77

77 revealed that 39 (14 EI, 6 PM, 6 TF, and 13 OL) items have appreciable loadings (i.e., .30 ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the fac tor it is intended to measure. The remaining 24 (5 EI, 8 PM, 2 TF and 9 OL) items did not have subst antial factor loadings (i.e., < .30 ) on their intended factors and conversely display high residual variances. Each of these items accounted for a signific ant proportion of variance that is not explained by its associated factor. Most factor correlations are significant and range from low (r = .077 for PM and TF) to moderate (r = .439 for PM and OL). This suggests that factor correlations for the Mongoli distinct on the constructs they measure. A review of the initial CFA solution indicated significant misspecifications of factorial structure for items 1, 7, 11, 16, 18, 23 31, 33, 35, 39, 40, 49, 50, 52, and 59. Allowing some of these items to cross load on other latent factors may improve the model fit statistics. After evaluating the content of these items, items 1, 31, 33, and 59 appeared to have substantive and logical meaning to other factors other than the factors they are associated to measure. Consequently, the aforementioned items were allowed to cross load: item 59 on EI, item 33 on TF, and items 1 and 31 on OL. The resulting model fit statistics of the modified C FA solution for the Mongo lia data yielded a significant 2 (1880) = 3078.819, p < .001; RMSEA = .025; CFI = .616; and TLI =.601. These fit statistics suggest that the data from Mongolian children did not exhibit a good fit to the four factor structure of t emperament, albeit with modifications. Table 3 14 reports the par ameter estimates of the four factor solution in the Mongolia data.

PAGE 78

78 Nigeria (N = 392 ) Response proportion of Nigerian children who endorsed specific temperament styles indicated their prefer ences for response options that measure practical more than imaginative and organized more than flexible styles. Approximately equal proportions of Nigeria children prefer extroverted and introverted as well as thinking and feeling styles. The initial CFA solution y ielded a significant 2 (1884) = 2282.416 p < .001; RMSEA = .023; CFI = .666; and TLI =.654. As a whole, these fit statistics suggest that the data from Nigeria children did not exhibit a good fit to the four factor structure of temperament. An inspecti on of the standardized factor loadings of item indicators revealed that 50 items (13 EI, 7 PM, 5 TF, and 15 OL items) have appreciable loadings (i.e., ) to their associated factors. Each of these items contributed a significant amount of vari ance that can be explained by the factor it is intended to measure. The remaining 23 items (6 EI, 7 PM, 3 TF, and 7 OL items) did not have signi ficant factor loadings (i.e., < .30 ) to their intended factors and conversely have high residual variances. Eac h of these items accounted for a significant proportion of variance that is not explained by its associated factor. Most factor correlations are significant and range from low (r = .027 for EI and OL) to moderate (r = .635 for PM and OL), with no eviden .85). Although most factors are related, they do not overlap and appear to be distinct in terms of the construct they measure. A review of the initial CFA solution indicated a significant misspecification of factorial st ructure for items 16, 52, 62, and 67. Allowing these items to cross load on other factors may improve the model fit statistics. However, only items 52 and 62

PAGE 79

79 appear to have substantive and logical meaning associated with OL and TF, respectively. When item 52 and 62 were allowed to cross load on OL and TF, the resulting model fit statistics of the modified CFA solution yielded a significant 2 (1882) = 2249.523, p < .001; RMSEA = .022; CFI = .692; and TLI =.681. These fit statistics suggest that the data from Nigerian children did not exhibit a good fit to the four factor structure of temperament, alb eit with modifications. Table 3 15 reports the para meter estimates of the four factor solution in the Nigeria data. Pakistan (N = 458) Pakistani children prefer practical more than imaginative, feeling more than thinking, and organized more than flexible styles. Approximately equal proportion of Pakistan children prefers extroverted and introverted styles. The initial CFA solution yielded a significant 2 (1884) = 2644.864, p < .001; RMSEA = .030; CFI = .504; and TLI =.486. As a whole, these fit statistics suggest that the data from Pakistani children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized factor loadings of item indicators revealed that 36 items (10 EI, 8 PM, 5 TF, and 13 OL items) have appreciable loadings (i.e., ) to their associated factors. Each of these items contributed a significant amount of variance that can be explained by t he factor it is inten ded to measure. The remaining 27 items (9 EI, 6 PM, 3 TF, and 9 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted fo r a significant proportion of variance that is not explained by its associated factor.

PAGE 80

80 Except for a moderate and significant correlation between PM and OL (r = .681), all other factor correlations are low and not significant. Multicollinearity (i.e., r not evident, thus suggesting that the factors are distinct in the constructs they measure. A review of the initial CFA solution indicated a significant misspecification of factorial structure for items 7, 11, 13, 18, 19, 27, 40, 49, 55, 57, 62, and 67. Allowing these items to cross load on other factors may improve the model fit statistics. However, only items 27 and 40 seemed to have substantive and logical meaning associated with PM and EI, respectively. When items 27 and 40 were allowed to cr oss load on PM and EI the model fit statistics of the modified CFA solution yielded a significant 2 (1882) = 2619.891, p < .001; RMSEA = .029; CFI = .519; and TLI =.501. These fit statistics suggest that the data from Pakistani children did not exhibit a good fit to the four factor structure of temperament as measured by the SSQ, albeit with modificati ons. Table 3 16 reports the par ameter estimates of the four factor solution in the Pakistan data. Philippines (N = 399) Filipino children prefer practical more than imaginative and organized more than flexible styles. Approximately equal proportions of Filipino children prefer extroverted and introverted as well as thinking and feeling styles. The initial CFA solution yielded a significant 2 (1884) = 2357.626, p < .001; RMSEA = .025; CFI = .678; and TLI =.666. As a whole, these fit statistics suggest that the data from Filipino children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized factor loadings of item indicators revealed that 36 items (12 EI, 6 PM, 3 TF, and 15 OL items) h ave appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by th e factor it is intended to measure. The

PAGE 81

81 remaining 27 items (7 EI, 8 PM, 5 TF, and 7 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversel y have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor. All factor correlations are significant and gener ally in the low range (i.e., r .300), except for a moderate correlation between PM and OL (r = .500). distinct on the constructs they measure. A review of the initial CFA solution indicated a significant misspecification of factorial structure for items 1, 4, 40, 48, and 67. Allowing these items to cross load on other latent factors may improve the model fit statistics. However, only items 1 and 40 seemed to have substantive and logical meaning associated with OL and EI, respectively. When items 1 and 40 were allowed to cross load on the OL and EI factors, respectively, the resulting model fit statistics of the modified CFA solution yielded a significant 2 (1882) = 2300.994, p < .001; RMSEA = .024; CFI = .715; and TLI =.704. These fit statistics suggest that the data from Filipino children did not exhibit a good fit to the four factor structure of temperament, alb eit with modifications. Table 3 17 reports the parameter estimates of the four factor model in the Philippine data. Poland (N = 434) Polish children prefer introverted more than extroverted, practical more than imaginative, and organized more than flexible styles. Approximately equal proportions of Polish children prefer thinking and feeling styles. The initial CFA solution yielded a significant 2 (1884) = 2792.991, p < .001; RMSEA = .033; CFI = .620; and TLI =.606. As a whole, these fit statistics suggest that

PAGE 82

82 the data from Polish children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized fa ctor loadings of item indicators revealed that 44 items (13 EI, 9 PM, 6 TF, and 16 OL items) have appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the factor it is intended to measure. The remaining 19 items (6 EI, 5 PM, 2 TF, and 6 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor. Most factor correlations ar e significant and low (i.e., r moderate correlation between PM and OL (r = not evident. Although the factors generally are related with one another, they reflect distinct constructs. A review of the initial CFA soluti on indicated a significant misspecification of factorial structure for items 7, 8, 22, 28, 33, 40, 45, 50, 59, 62, 64, and 68. Allowing these items to cross load on other latent factors would likely improve the model fit statistics. However, only items 8, 28, 40, 50, and 59 seem to have substantive and logical meaning associated with EI. When those five items were allowed to cross load on EI, the resulting model fit statistics of the modified CFA solution yielded a significant 2 (1879) = 2687.857, p < .001; RMSEA = .031; CFI = .662; and TLI =.648. These fit statistics suggest that the data from Polish children did not exhibit a good fit to the four factor structure of temperament as measured by the SSQ, albeit with modifications

PAGE 83

83 Table 3 18 reports the par ameter estimates of the four factor solution in the Poland data. Romania (N = 391) Romanian children prefer extroverted more than introverted, practical more than imaginative, thinking more than feeling and organized more than flexible styles. The initial CFA solution yielded a significant 2 (1884) = 2650.835, p < .001; RMSEA = .032; CFI = .654; and TLI =.641. As a whole, these fit statistics suggest that the data from Filipino children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized factor loadings of item indicators revealed that 45 items (12 EI, 9 PM, 6 TF, and 18 OL items) h ave appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by th e factor it is intended to measure. The remaining 18 items (7 EI, 5 PM, 2 TF, and 4 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explained by its associated factor. Most factor correlations ar e significant and low (i.e., r moderate correlation between PM and OL (r = not evident. Although the factors generally are related with one another, they are distinct in the constructs they measure. A review of the ini tial CFA solution indicated a significant misspecification of factorial structure for items 1, 16, 33, 40, 43, 52, 62, and 67. Allowing these items to cross load on other latent factors may improve the model fit statistics. However, only

PAGE 84

84 items 1 and 40 see med to have substantive and logical meaning associated with OL and EI factors, respectively. When these items were allowed to cross load, the resulting model fit statistics of the modified CFA solution yielded a significant 2 (1882) = 2595.766, p < .001; RMSEA = .031; CFI = .678; and TLI =.666. These fit statistics suggest that the data from Romanian children did not exhibit a good fit to the four factor structure of temperament, alb eit with modifications. Table 3 19 reports the parameter estimates of the four factor solution in the Romania data. Samoa (N = 400) Samoan children prefer extroverted more than introverted, practical more than imaginative, thinking more than feeling and organized more than flexible styles. The initial CFA solution yielded a significant 2 (1884) = 2501.372, p < .001; RMSEA = .029; CFI = .620; and TLI =.606. As a whole, these fit statistics suggest that the data from Samoa children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized fac tor loadings of item indicators revealed that 39 items (13 EI, 8 PM, 4 TF, and 14 OL items) h ave appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the f actor it is intended to measure. The remaining 24 items (6 EI, 6 PM, 4 TF, and 8 OL items) did not have signif icant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted for a significant proportion of variance that is not explai ned by its associated factor. All factor correlations are high, ranging from r = .713 (EI and TF) to r = .959 (TF .863), PM and OL (r

PAGE 85

85 = .857), and TF and OL (r = .959). These correlations imply that the factor s overlap and may share similar meanings on what they measure. A review of the initial CFA solution indicated a significant misspecification of factorial structure for item 52; that is, allowing item 52 to cross load on TF and OL traits may improve the mo del fit statistics. However, the content meaning of item 52 is not substantively and logically related to TF and OL. Hence, there was no compelling reason to pursue a modificati on of the CFA solution. Table 3 20 reports the p arameter estimates of the four factor solution in the Samoa data. Singapore (N = 483) Singaporean children prefer practical more than imaginative, feeling more than thinking and organized more than flexible styles. Approximately equal proportions of Singaporean children prefer extrover ted and introverted styles. Initial CFA solution yielded a significant 2 (1884) = 2463.218, p < .001; RMSEA = .025; CFI = .732; and TLI =.725. As a whole, these fit statistics suggest that the data from Singaporean children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardiz ed factor loadings of item indicators revealed that 49 (16 EI, 8 PM, 6 TF, and 19 OL) items h ave appreciable loadings (i.e., .30 ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the f actor it is intended to measure. The remaining 14 (3 EI, 6 PM, 2 TF, and 3 OL) items did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted for a signif icant proportion of variance that is not explained by its associated factor.

PAGE 86

86 constructs. A review of the initial CFA solution indicated a significant misspec ification of factorial structure for items 17, 25, 43, 52, 59, 68, and 69. Allowing these items to cross load on other factors may improve the model fit statistics. Among the misspecified items, only item 59 seemed to have substantive and logical meaning a ssociated with EI. When item 59 was allowed to cross load on EI, the resulting model fit statistics of the modified CFA solution yielded a significa nt 2 (1883) = 2432.524, p < .001; RMSEA = .025; CFI = .748; and TLI =.739. These fit statistics suggest that the data from Singapore children did not exhibit a good fit to the four factor structure of temperament, alb eit with modifications. Table 3 21 reports the par ameter estimates of the four factor solution in the Singapore data. U.S. (N = 7,902) Children from the U.S prefer imaginative more than practical and organized more than and flexible styles. Approximately equal proportions of U.S. children prefer extroverted and introverted as well as thinking and feeling styles. The initial CFA solution yielded a significant 2 (1884) = 2488.575, p < .001; RMSEA = .025; CFI = .759; and TLI =.750. As a whole, these fit statistics suggest that the data from U.S. children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized fac tor loadings of item indicators revealed that 54 items (14 EI, 11 PM, 7 TF, and 22 OL items), including all OL have appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the factor it is

PAGE 87

87 intended to measure. The remaining 8 items (5 EI, 3 PM, and 1 TF items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items acco unted for a significant proportion of variance that is not explained by its associated factor. Factor correlations generally are low, ranging from r = .048 for EI and PM to r = seemingly measure distinct. A review of the initial CFA solution indicated a significa nt misspecification of factorial structure for items 1, 16, 26, 42, 55, 62, and 65. Allowing these items to cross load on other latent factors may improve the model fit statistics. After evaluating the content of these items, items 1, 16, 26, 52, and 62 ap peared to have substantive and logical meaning to factors other than thos e they are intended to measure. Consequently, the aforementioned items were allowed to cross load: items 16, 26, and 52 on TF, and items 1 and 62 on OL. The resulting model fit statis tics of the modified CFA solution yielded a significant 2 (1880) = 2407.145, p < .001; RMSEA = .024; CFI = .789; and TLI =.781. These fit statistics suggest that the data from U.S. children did not exhibit a good fit to the four factor structure of temper ament, albeit with modifications. Table 3 22 reports the par ameter estimates of the four factor solution in the U.S. data. Venezuela (N = 411) Venezuelan children prefer extroverted more than introverted, practical more than imaginative, and thinking mor e than feeling styles. Approximately equal proportion of Venezuelan children prefer organized and flexible styles.

PAGE 88

88 The initial CFA solution yielded a significant 2 (1884) = 2586.444, p < .001; RMSEA = .030 ; CFI = .648; and TLI =.635. These fit statistics suggest that the data from Venezuelan children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardized factor l oadings of item indicators revealed that 40 items (15 EI, 4 PM, 5 TF, and 16 OL items) have appreciable loadings (i.e., ) on their associated factors. Each of these items contributed a significant amount of variance that can be explained by the factor it is inten ded to measure. The remaining 23 items (4 EI, 10 PM, 3 TF, and 6 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted for a sign ificant proportion of variance that is not explained by its associated factor. Most factor correlat ions are in the low range (i.e., r correlation between PM and OL (r = evident, the factors seemingly measure distinct constructs. A review of the initial CFA solution indicated a significant misspecification o f factorial structure for items 1, 7, 18, 39, 59, 62, 67, and 68. Allowing these items to cross load on other latent factors improve the model fit statistics. Among the misspecified items, only items 1 and 59 seemed to have substantive and logical meaning associated with OL and EI, respectively. When items 1 and 59 were allowed to cross load on OL and EI factors, the resulting model fit statistics of the modified CFA solution yielded a significant 2 (1882) = 2517.479, p < .001; RMSEA = .029; CFI = .682; and TLI =.670. These fit statistics suggest that the data from Venezuelan children did not exhibit a good fit to the four factor structure of temperament, al beit with modification.

PAGE 89

89 Table 3 23 reports the par ameter estimates of the four factor solution in the Venezuela data. Zimbabwe (N = 490) Zimbabwe children prefer extroverted more than introverted, practical more than imaginative, feeling more than thinking, and organized more than flexible styl es. The initial CFA solution yielded a significant 2 (1884) = 2300.126, p < .001, RMSEA = .021 CFI = .664, and TLI =.652. As a whole, these fit statistics suggest that the data from Zimbabwe children did not exhibit a good fit to the four factor structure of temperame nt. An inspection of the standardiz ed factor loadings of item indicators revealed that only 30 (4 EI, 7 PM, 3 TF, and 17 OL) out of 63 items have appreciable loadings (i.e., ) on their associated factors. Each of these 31 items contributed a significant amount of variance that can be e xplained by the fact or it measures. The remaining 33 items (15 EI, 8 PM, 5 TF, and 5 OL items) did not have signi ficant factor loadings (i.e., < .30 ) on their associated factors and conversely have high residual variances. Each of these items accounted fo r a significant proportion of variance that is not explained by its associated latent factor. Most factor correlations are significant and range from low (r = .183 for EI and TF) to high (r = .763 for PM and OL). As multicollinearity factors seemingly measure distinct constructs. A review of the initial CFA solution indicated a significant misspecification of factorial structure for items 7, 12, 26, 46, and 52. Allowing these items to cross load on othe r factors improve the model fit statistics. Among the misspecified items, only item 52 seemed to have substantive and logical meaning associated with OL trait. When

PAGE 90

90 item 52 was allowed to cross load on OL trait, the resulting model fit statistics of the mo dified CFA solution yielded a significant 2 (1883) = 2281.020, p < .001; RMSEA = .021; CFI = .679; and TLI =.667. These fit statistics suggest that the data from Zimbabwe children did not exhibit a good fit to the four factor structure of temperament, al beit with modification. Table 3 24 reports t he par ameter estimates of the four factor solution in the Zimbabwe data

PAGE 91

91 Table 3 1. Participants per country with at least 6 item responses missing Country Total Sample Missing Item Responses Actual Sample Size Australia 369 1 368 Brazil 461 0 461 China 400 1 399 Costa Rica 432 1 431 Egypt 954 34 920 Gaza 400 0 400 Hungary 401 0 400 Iran 511 1 510 Israel 253 35 218 Japan 493 0 493 Mongolia 1009 6 1003 Nigeria 400 8 392 Pakistan 463 5 458 Philippines 400 1 399 Poland 440 6 434 Romania 391 0 391 Samoa 400 0 400 Singapore 483 0 483 United States 7,902 0 7,902 Venezuela 411 0 411 Zimbabwe 492 2 490

PAGE 92

92 Table 3 2. Cronbach alpha coefficients of the temperament traits in 21 countries Country EI (19 items) PM (14 items) TF (8 items) OL (22 items) 95% CI 95% CI 95 %CI 95% CI Australia .70 .66 .75 .56 .49 .63 .64 .58 .70 .80 .77 .83 Brazil .72 .68 .76 .52 .45 .58 .55 .48 .61 .46 .39 .53 China .73 .69 .77 .55 .48 .61 .39 .30 .48 .73 69 .76 Costa Rica .71 .66 .75 .36 .26 .44 .41 .32 .49 .75 .71 .78 Egypt .49 .44 .54 .33 .26 .39 .29 .22 .36 .34 .28 .40 Gaza .61 .55 .66 .41 .32 .50 .38 .28 .47 .72 .69 .76 Hungary .65 .60 .70 .48 .40 .55 .30 .20 .40 .55 .48 .61 Iran .60 .55 .65 .31 22 .39 .12 .01 .24 .76 .73 .79 Israel .75 .68 .80 .56 .46 66 .62 .53 .69 .65 .58 .72 Japan .78 .75 .81 .51 .44 .57 .57 .51 .62 .73 .69 .76 Mongolia .54 .50 .59 .23 .16 .30 .41 .36 .47 .43 .38 .49 Nigeria .62 .56 .67 .47 .39 .55 .39 .30 .48 .66 .60 .70 Pakistan .56 .49 .62 .48 .41 .55 .39 .30 .47 .60 .54 .65 Philippines .62 .57 .67 .42 .33 .50 .34 .24 .44 .61 .57 .67 Poland .62 .57 .67 .54 .47 .60 .58 .52 .64 .71 .67 .75 Romania .67 .62 .72 .49 .40 .55 .41 .32 .50 .76 .73 .80 Samoa .56 .50 .62 .32 22 .41 .21 .09 .32 .35 .25 .43 Singapore .73 .69 .76 .49 .42 .55 .49 .42 .56 .69 .65 .73 United States .69 .68 .70 .55 .54 .57 .35 .32 .37 .68 .67 .69 Venezuela .62 .57 .68 .31 .21 .41 .33 .22 .42 .66 .61 .71 Zimbabwe .51 .45 .57 .25 .15 .35 .20 .09 31 .60 .54 .65 .64 .44 .40 .63

PAGE 93

93 Table 3 3 Overall goodness of fit indices for individual c ountry CFA Country WLSMV 2 df RMSEA (90% C.I.) CFI TLI Australia Initial solution 2606.560 1884 .032 (.029 .035) .733 .723 Modif ied solution 2548.690 1881 .031 (.028 .034) .753 .734 Brazil Initial solution 2990.948 1884 .033 (.031 .035) .639 .626 Modified solution 2893.080 1881 .033 (.031 .035) .670 .658 China 2754.924 1884 .034 (.031 .037) .741 .731 Initial solut ion Costa Rica Initial solution 3164.511 1884 .026 (.023 .029) .590 .575 Modified solution 3088.953 1880 .025 (.022 .028) .613 .598 Egypt Initial solution 3146.025 1884 .027 (.025 .029) .513 .495 Gaza Initial solution 2 461.500 1884 .029 (.026 .032) .755 .746 Hungary Initial solution 2617.849 1884 .031 (.028 .034) .475 .456 Modified solution 2592.397 1882 .031 (.028 .034) .492 .472 Iran Initial solution 2878.989 1884 .032 (.030 .035) .722 .712 Modi fied solution 2866.009 1883 .032 (.030 .034) .725 .715 Israel Initial solution 2175.017 1884 .027 (.025 .029) .749 .739 Modified solution 2144.574 1882 .025 (.022 .028) .773 .765 Japan Initial solution 2904.025 1884 .033 (.031 .035) .732 .722 Modified solution 2861.529 1881 .033 (.031 .035) .743 .733 Mongolia Initial solution 3166.944 1884 .026 (.024 .028) .589 .574 Modified solution 3078.819 1880 .025 (.022 .028) .616 .601 Nigeria Initial solution 2282.416 1884 .023 (.020 .027) .666 .654 Modified solution 2249.523 1882 .022 (.018 .026) .692 .681 Pakistan Initial solution 2644.864 1884 .030 (.027 .032) .504 .486 Modified solution 2619.891 1882 .029 (.026 .032) .519 .501 Philippines Initial solution 2357.626 1884 .025 (.022 .028) .678 .666 Modified solution 2300.994 1882 .024 (.021 .027) .715 .704 Poland Initial solution 2792.991 1884 .033 (.031 .036) .620 .606 Modified solution 2687.857 1879 .031 (.028 .034) .662 .648

PAGE 94

94 Tab le 3 3. Continued. Country WLSMV 2 df RMSEA (90% C.I.) CFI TLI Romania Initial solution 2650.835 1884 .032 (.029 .035) .654 .641 Modified solution 2595.766 1882 .031 (.028 .034) .678 .666 Samoa Initial solution 2501.372 1884 .029 (.026 .032) .6 33 .6 19 Singapore Initial solution 2463.218 1884 .025 (.022 .028) .732 .725 Modified solution 2432.524 1883 .025 (.022 .028) .748 .739 United States Initial solution 2488.575 1884 .025 (.023 .028) .759 .750 Modified solution 2407.145 1880 .024 (.021 .027) .789 .781 Venezuela Initial solution 2586.444 1884 .030 (.027 .033) .648 .635 Modified solution 2517.479 1882 .029 (.026 .032) .682 .670 Zimbabwe Initial solution 2300.126 1884 .021 (.018 .024) .664 .652 Modified solution 2281.020 18 83 .021 (.018 .024) .679 .667

PAGE 95

95 Table 3 4. Parameter estimates of the four factor model in Australia data Trait/Item # Factor loading Estimate/SE Residual Variance Threshold EI 4 0.501 7.300 0.749 0.213 7 0.494 7.011 0.756 0.705 13 0.308 3.97 2 0.905 0.525 16 0.389 5.704 0.849 0.225 19 0.513 7.799 0.737 0.287 22 0.376 5.436 0.859 0.164 26 0.376 5.268 0.859 0.092 33 0.623 10.605 0.612 0.371 36 0.442 6.211 0.805 0.235 39 0.202 2.630 0.959 0.296 40 0.541 8.577 0.707 0.214 42 0.633 11 .481 0.599 0.248 46 0.575 9.071 0.669 0.371 49 0.498 7.614 0.752 0.075 52 0.426 6.165 0.819 0.075 57 0.584 8.891 0.658 0.707 62 0.206 2.754 0.957 0.114 65 0.301 3.832 0.910 0.438 67 0.393 5.747 0.846 0.038 PM 3 0.451 5.680 0.797 0.221 6 0.527 6.450 0.722 0.359 9 0.228 2.759 0.948 0.137 11 0.037 0.398 0.999 0.531 15 0.401 5.014 0.839 0.068 21 0.565 7.586 0.681 0.151 25 0.443 5.745 0.804 0.221 31 0.425 5.344 0.820 0.352 34 0.384 4.713 0.852 0.206 45 0.202 2.351 0.959 0.052 48 0.581 7.401 0.662 0.486 51 0.220 2.621 0.952 0.249 64 0.548 7.226 0.700 0.294 68 0.225 2.614 0.949 0.231

PAGE 96

96 Table 3 4. Continued Trait/Item # Factor loading Estimate/SE Residual Variance Threshold TF 12 0.582 8.836 0.661 0.021 18 0.730 11.752 0.467 0.096 28 0.742 11.894 0.450 0.130 37 0.702 11.016 0.507 0.565 43 0.213 2.493 0.955 0.338 50 0.383 4.747 0.853 0.330 55 0.438 5.441 0.808 0.939 61 0.572 8.299 0.673 0.289 OL 2 0.391 6.039 0.847 0.178 5 0.599 11.535 0.641 0.034 8 0.22 0 2.633 0.952 0.918 14 0.278 3.549 0.923 0.794 17 0.718 16.285 0.484 0.144 20 0.674 13.758 0.545 0.481 23 0.506 8.505 0.744 0.106 27 0.453 7.335 0.794 0.106 32 0.709 14.993 0.497 0.438 35 0.396 5.782 0.843 0.698 38 0.655 13.409 0.570 0.116 40 0 .340 4.855 0.884 0.393 44 0.293 4.130 0.914 0.403 47 0.484 8.139 0.766 0.456 53 0.396 6.245 0.843 0.106 56 0.375 5.306 0.859 0.541 58 0.650 13.190 0.577 0.182 59 0.509 8.720 0.741 0.113 60 0.788 20.781 0.379 0.158 63 0.313 4.391 0.902 0.446 66 0.651 13.869 0.576 0.082 69 0.596 10.881 0.644 0.248

PAGE 97

97 Table 3 5. Parameter estimates of the four factor model in Brazil data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.453 8.422 0.795 0.003 4 0.594 12.054 0 .647 0.382 7 0.376 6.440 0.858 0.216 13 0.422 7.964 0.822 0.194 16 0.619 13.556 0.617 0.205 19 0.568 12.158 0.678 0.112 22 0.558 10.959 0.688 0.399 26 0.640 14.474 0.590 0.128 33 0.365 5.930 0.867 0.520 36 0.466 7.908 0.782 0.558 39 0.510 10.0 64 0.740 0.014 42 0.501 7.892 0.749 0.704 46 0.708 18.018 0.499 0.101 49 0.532 10.481 0.717 0.222 52 0.420 7.727 0.824 0.079 57 0.630 12.049 0.603 0.649 62 0.429 7.469 0.816 0.358 65 0.261 3.885 0.932 0.732 67 0.428 7.618 0.817 0.284 PM 3 0.311 4.842 0.903 0.596 6 0.341 5.123 0.884 0.364 9 0.809 17.478 0.346 0.683 11 0.142 2.081 0.980 0.603 15 0.527 9.364 0.722 0.183 21 0.231 3.388 0.947 0.222 25 0.653 14.457 0.574 0.041 31 0.276 4.274 0.924 0.656 34 0.335 5.516 0.888 0.483 45 0.303 4.596 0.908 0.189 48 0.328 5.293 0.892 0.352 51 0.793 19.776 0.372 0.447 64 0.242 3.584 0.941 0.228 68 0.566 10.316 0.679 0.520

PAGE 98

98 Table 3 5. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.623 10.48 0 0.612 0.483 18 0.746 12.396 0.443 0.435 28 0.453 6.868 0.795 0.014 37 0.426 6.393 0.819 0.284 43 0.425 6.262 0.820 0.465 50 0.395 6.027 0.844 0.019 55 0.560 8.190 0.687 0.718 61 0.505 8.234 0.745 0.189 OL 2 0.322 4.831 0.896 0.435 5 0.360 5.750 0.870 0.057 8 0.511 8.327 0.739 0.539 14 0.428 6.086 0.817 1.047 17 0.756 16.714 0.429 0.539 20 0.397 6.024 0.842 0.489 23 0.380 5.668 0.856 0.636 27 0.143 2.039 0.980 0.429 32 0.609 11.135 0.629 0.453 35 0.476 7.369 0.773 0.797 38 0.521 9.137 0.729 0.211 40 0.263 4.036 0.931 0.347 44 0.105 1.595 0.989 0.312 47 0.258 3.974 0.934 0.150 53 0.017 0.234 1.000 0.558 56 0.147 1.952 0.979 0.813 58 0.745 15.332 0.445 0.739 59 0.070 1.042 0.995 0.095 60 0.683 14.050 0.533 0.387 63 0.202 3.070 0.959 0.256 66 0.721 16.024 0.480 0.417 69 0.497 8.570 0.753 0.228

PAGE 99

99 Table 3 .30 factor loadings by factor solution for China data Trait No. of items 2 factor model 3 factor model 4 factor model 5 factor model 1 2 1 2 3 1 2 3 4 1 2 3 4 5 EI 19 1 16 1 1 16 17 1 3 2 1 18 3 PM 14 3 1 3 10 3 13 5 12 TF 8 1 1 1 3 1 4 1 4 OL 22 18 1 17 3 1 18 1 5 20 1 1 5 Model Fit Indices 2 2524.415 2288.786 2123. 139 1987.430 df 1828 1767 1707 1640 p .000 .000 .000 .000 CFI .793 .846 .876 .899 TLI .779 .880 .858 .880 RMSEA .031 .027 .025 .023

PAGE 100

100 Table 3 7. Parameter estimates of the four factor model in Costa Rica data Trait/Item Factor loading Estimate/SE R esidual Variance Threshold EI 1 0.216 4.711 0.911 0.084 4 0.555 12.216 0.784 0.957 7 0.054 0.980 0.931 0.450 13 0.439 9.966 0.747 0.748 16 0.540 12.425 0.637 0.170 19 0.474 11.447 0.707 0.372 22 0.333 7.200 0.845 0.146 26 0.557 14.515 0.56 0 0.123 33 0.310 6.970 0.958 0.088 36 0.377 8.618 0.950 0.281 39 0.450 10.840 0.673 0.579 42 0.490 11.154 0.781 0.586 46 0.430 10.022 0.645 0.304 49 0.404 8.844 0.846 0.348 52 0.256 5.634 0.724 0.203 57 0.623 14.846 0.531 0.935 62 0.449 10.5 53 0.776 0.518 65 0.208 4.384 0.955 0.586 67 0.264 5.740 0.833 0.372 PM 3 0.075 1.052 0.910 0.513 6 0.010 0.150 0.973 0.480 9 0.577 7.442 0.988 0.717 11 0.186 2.949 0.985 1.116 15 0.150 2.327 0.989 0.012 21 0.391 5.801 0.982 0.023 25 0.300 4.672 0.992 0.052 31 0.034 0.516 0.890 0.801 34 0.336 5.045 0.999 0.296 45 0.191 3.017 0.875 0.059 48 0.376 5.863 0.594 0.622 51 0.446 6.921 1.000 0.600 64 0.026 0.412 0.922 0.056 68 0.147 2.243 0.910 0.809

PAGE 101

101 Table 3 7. Continued Trait /Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.527 9.225 0.723 0.241 18 0.373 6.053 0.739 0.259 28 0.424 7.448 0.969 0.131 37 0.478 8.100 0.876 1.427 43 0.082 1.409 0.995 0.353 50 0.132 2.252 0.958 0.247 55 0.408 6.80 9 0.975 0.867 61 0.621 10.788 0.974 0.143 OL 2 0.341 7.042 0.925 0.310 5 0.412 8.725 0.706 0.420 8 0.121 2.331 0.939 0.926 14 0.367 6.550 0.783 1.729 17 0.339 6.413 0.281 0.982 20 0.482 10.431 0.614 0.433 23 0.330 6.642 0.813 1.217 27 0.248 4.996 0.897 0.143 32 0.568 10.009 0.445 0.944 35 0.011 0.224 0.847 1.814 38 0.503 10.730 0.606 0.669 40 0.074 1.443 0.968 0.372 44 0.270 5.327 0.723 0.268 47 0.004 0.076 0.769 0.152 53 0.377 6.896 0.762 0.363 56 0.195 3.882 0.682 0.190 58 0.365 7.401 0.652 0.755 59 0.488 10.604 1.000 0.061 60 0.415 8.561 0.364 0.676 63 0.269 5.396 0.969 0.642 66 0.487 9.150 0.455 0.738 69 0.178 3.355 0.681 0.410

PAGE 102

102 Table 3 factor solu tion for Egypt data Trait No. of items 2 factor model 3 factor model 4 factor model 5 factor model 1 2 1 2 3 1 2 3 4 1 2 3 4 5 EI 19 1 2 2 2 9 2 9 2 2 8 PM 14 4 1 5 1 3 5 1 1 5 2 2 2 TF 8 1 2 1 2 1 1 2 1 1 2 OL 22 10 5 11 6 6 5 4 8 5 6 Model Fit Indices 2 2795.259 2496.157 2235.302 2062.220 df 1828 1767 1707 1648 p .000 .000 .000 .000 CFI .627 .718 .796 .840 TLI .601 .689 .767 .810 RMSEA .024 .021 .018 .017

PAGE 103

103 Table 3 factor solution for Gaza data Trait No. of items 2 factor model 3 factor model 4 factor model 5 factor model 1 2 1 2 3 1 2 3 4 1 2 3 4 5 EI 19 3 10 4 9 5 3 9 4 3 5 5 1 7 PM 14 8 4 11 4 6 8 4 5 8 5 5 1 TF 8 3 3 2 4 2 4 3 2 4 1 OL 22 19 1 18 1 19 1 1 1 16 2 1 2 1 Model Fit Indices 2 2152.565 2037.012 1932.922 1834.679 df 1828 1767 1707 1648 p .000 .000 .000 .000 CFI .862 .885 .904 .921 TLI .853 .873 .890 .906 RMSEA .022 .020 .019 .017

PAGE 104

104 Table 3 10. Parameter estimates of the four factor model in Hungary data Trait/Item Fac tor loading Estimate/SE Residual Variance Threshold EI 1 0.463 7.288 0.785 0.091 4 0.574 8.379 0.671 0.708 7 0.172 2.167 0.970 0.741 13 0.468 7.714 0.781 0.128 16 0.321 4.630 0.897 0.016 19 0.360 5.163 0.870 0.224 22 0.300 3.968 0.911 0.50 5 26 0.333 4.862 0.889 0.022 33 0.197 2.569 0.961 0.361 36 0.395 5.842 0.844 0.282 39 0.591 10.089 0.651 0.041 42 0.458 6.603 0.790 0.519 46 0.342 5.085 0.883 0.047 49 0.485 6.903 0.765 0.374 52 0.480 7.786 0.770 0.078 57 0.460 6.796 0.788 0.622 62 0.460 6.661 0.788 0.368 65 0.221 2.925 0.951 0.428 67 0.245 3.429 0.940 0.173 PM 3 0.321 4.390 0.897 0.749 6 0.531 8.071 0.718 0.122 9 0.010 0.121 1.000 0.835 11 0.340 4.809 0.884 0.415 15 0.243 3.357 0.941 0.128 21 0.222 3. 089 0.951 0.097 25 0.220 3.083 0.951 0.110 31 0.390 5.512 0.848 0.505 34 0.269 3.896 0.928 0.198 45 0.463 6.673 0.786 0.262 48 0.288 4.219 0.917 0.022 51 0.252 3.382 0.936 0.505 64 0.192 2.580 0.963 0.147 68 0.281 3.999 0.921 0.548

PAGE 105

105 Table 3 10. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.436 5.244 0.809 0.166 18 0.109 1.210 0.988 0.512 28 0.172 1.921 0.970 0.009 37 0.327 3.299 0.893 0.630 43 0.588 6.338 0.654 0.230 50 0.175 1.958 0.969 0.1 54 55 0.105 1.004 0.989 0.880 61 0.427 5.107 0.818 0.205 OL 2 0.255 3.836 0.935 0.072 5 0.655 12.391 0.571 0.456 8 0.403 5.536 0.838 0.645 14 0.255 3.826 0.935 0.097 17 0.615 9.898 0.622 0.630 20 0.013 0.186 1.000 0.003 23 0.398 5.760 0.842 0.630 27 0.053 0.703 0.997 0.676 32 0.466 7.086 0.783 0.463 35 0.083 1.039 0.993 0.638 38 0.230 3.228 0.947 0.394 40 0.313 3.962 0.902 0.700 44 0.305 4.392 0.907 0.321 47 0.304 4.656 0.908 0.224 53 0.220 3.084 0.951 0.308 56 0.287 3.519 0.91 8 0.791 58 0.533 8.828 0.716 0.078 59 0.096 1.223 0.991 0.669 60 0.700 12.135 0.510 0.615 63 0.387 5.926 0.850 0.003 66 0.456 6.628 0.792 0.519 69 0.203 2.838 0.959 0.301

PAGE 106

106 Table 3 11. Parameter estimates of the four factor model in Iran data Trait/ Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.417 6.830 0.826 0.485 4 0.115 1.739 0.987 0.425 7 0.324 5.189 0.895 0.422 13 0.117 1.804 0.986 0.064 16 0.278 4.366 0.923 0.238 19 0.031 0.484 0.999 0.054 22 0.419 6.838 0. 825 0.377 26 0.045 0.695 0.998 0.148 33 0.026 0.388 0.999 0.146 36 0.190 2.970 0.964 0.325 39 0.291 4.633 0.915 0.111 42 0.520 8.950 0.729 0.377 46 0.286 4.476 0.918 0.258 49 0.715 15.729 0.488 0.785 52 0.380 6.214 0.856 0.304 57 0.352 5 .889 0.876 0.279 62 0.218 3.373 0.952 0.356 65 0.416 6.935 0.827 0.325 67 0.233 3.623 0.946 0.027 PM 3 0.069 1.034 0.995 0.436 6 0.215 3.522 0.954 0.511 9 0.519 9.840 0.730 0.665 11 0.034 0.512 0.999 0.228 15 0.274 4.449 0.925 0.138 21 0.149 2.359 0.978 0.144 25 0.512 9.519 0.738 0.310 31 0.489 8.266 0.761 0.582 34 0.081 1.264 0.994 0.012 45 0.495 9.214 0.755 0.469 48 0.628 13.430 0.606 0.678 51 0.565 11.120 0.681 0.519 64 0.254 4.140 0.936 0.139 68 0.399 6.683 0.840 0. 404

PAGE 107

107 Table 3 11. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.007 0.083 1.000 0.000 18 0.224 2.830 0.950 0.335 28 0.179 2.215 0.968 0.074 37 0.322 4.382 0.897 0.108 43 0.135 1.668 0.982 0.153 50 0.981 8. 548 0.038 0.458 55 0.087 1.062 0.992 0.133 61 0.027 0.321 0.999 0.206 OL 2 0.358 6.296 0.872 0.424 5 0.409 7.189 0.833 0.372 8 0.319 5.451 0.898 0.218 14 0.080 1.298 0.994 0.208 17 0.658 15.221 0.567 0.653 20 0.459 8.705 0.789 0.218 23 0.5 71 12.244 0.674 0.623 27 0.058 0.942 0.997 0.123 32 0.690 16.680 0.524 0.547 35 0.340 5.885 0.885 0.069 38 0.491 9.553 0.759 0.469 40 0.518 10.167 0.732 0.297 44 0.458 8.735 0.790 0.208 47 0.579 12.042 0.665 0.223 53 0.535 10.560 0.714 0.346 56 0.502 9.588 0.748 0.541 58 0.593 12.995 0.648 0.627 59 0.627 14.113 0.607 0.617 60 0.691 17.363 0.523 0.617 63 0.197 3.301 0.961 0.143 66 0.645 14.994 0.583 0.528 69 0.467 8.270 0.782 0.528

PAGE 108

108 Table 3 12. Parameter estimates of the four factor model in Israel data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.279 3.087 0.922 0.037 4 0.763 12.682 0.418 0.552 7 0.628 8.483 0.606 0.522 13 0.534 6.907 0.715 0.491 16 0.434 4.728 0.811 0.576 19 0.530 7.001 0.719 0.0 47 22 0.459 4.340 0.789 1.248 26 0.551 7.463 0.696 0.355 33 0.203 2.276 0.959 0.128 36 0.448 4.740 0.799 0.678 39 0.540 6.945 0.709 0.422 42 0.658 8.113 0.567 0.734 46 0.708 10.458 0.498 0.660 49 0.470 5.252 0.780 0.667 52 0.478 5.873 0.772 0.486 57 0.693 9.892 0.520 0.674 62 0.297 3.390 0.912 0.122 65 0.285 2.827 0.919 0.594 67 0.466 5.808 0.783 0.094 PM 3 0.565 7.029 0.681 0.194 6 0.193 1.907 0.963 0.152 9 0.798 13.734 0.363 0.554 11 0.196 1.986 0.961 0.088 15 0.546 6.302 0.702 0.286 21 0.082 0.802 0.993 0.256 25 0.505 6.143 0.745 0.095 31 0.647 9.042 0.581 0.232 34 0.355 3.634 0.874 0.719 45 0.237 2.502 0.944 0.190 48 0.382 3.947 0.854 0.426 51 0.822 14.179 0.324 0.572 64 0.014 0.141 1.000 0.311 68 0.142 1 .240 0.980 0.715

PAGE 109

109 Table 3 12. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.551 6.716 0.697 0.282 18 0.752 10.859 0.434 0.220 28 0.710 10.020 0.496 0.046 37 0.536 5.648 0.713 0.435 43 0.068 0.631 0.995 0.4 35 50 0.479 4.951 0.771 0.253 55 0.589 6.370 0.653 0.530 61 0.632 8.349 0.600 0.006 OL 2 0.400 4.555 0.840 0.146 5 0.518 6.942 0.732 0.285 8 0.553 6.519 0.694 0.796 14 0.117 1.151 0.986 0.538 17 0.777 13.503 0.396 0.625 20 0.617 8.627 0.620 0.439 23 0.448 5.172 0.799 0.401 27 0.092 0.825 0.992 0.617 32 0.627 8.888 0.607 0.761 35 0.216 2.102 0.953 0.542 38 0.748 12.095 0.440 0.197 40 0.340 3.898 0.884 0.205 44 0.317 3.542 0.899 0.270 47 0.031 0.280 0.999 0.812 53 0.237 2.592 0. 944 0.133 56 0.246 2.239 0.940 0.902 58 0.510 6.420 0.740 0.328 59 0.019 0.196 1.000 0.190 60 0.732 12.322 0.464 0.469 63 0.176 1.407 0.969 1.248 66 0.752 13.101 0.434 0.552 69 0.584 7.925 0.658 0.490

PAGE 110

110 Table 3 13. Parameter estimates of the four factor model in Japan data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.646 14.318 0.583 0.388 4 0.664 13.900 0.559 0.651 7 0.290 5.022 0.916 0.421 13 0.518 10.121 0.732 0.449 16 0.661 15.877 0.563 0.334 19 0.466 7.734 0.783 0.832 22 0.577 12.108 0.667 0.302 26 0.606 13.856 0.632 0.074 33 0.437 7.564 0.809 0.489 36 0.370 6.046 0.863 0.524 39 0.504 10.225 0.746 0.059 42 0.550 8.462 0.698 1.002 46 0.638 15.217 0.593 0.223 49 0.588 12.651 0.655 0 .140 52 0.534 10.754 0.715 0.223 57 0.686 15.673 0.529 0.589 62 0.387 6.692 0.851 0.472 65 0.380 6.901 0.856 0.156 67 0.570 12.347 0.675 0.275 PM 3 0.178 2.530 0.968 0.003 6 0.057 0.795 0.997 0.028 9 0.753 13.457 0.433 0.768 11 0.383 5.638 0.854 0.218 15 0.377 5.438 0.858 0.318 21 0.200 2.709 0.960 0.421 25 0.573 9.969 0.672 0.244 31 0.096 1.336 0.991 0.115 34 0.129 1.797 0.983 0.583 45 0.342 4.859 0.883 0.444 48 0.329 4.718 0.892 0.461 51 0.871 16.369 0.241 0.607 64 0.343 4.992 0.882 0.270 68 0.251 3.594 0.937 0.207

PAGE 111

111 Table 3 13. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.523 8.889 0.727 0.140 18 0.747 11.086 0.442 0.728 28 0.622 10.571 0.613 0.048 37 0.586 9.284 0. 656 0.161 43 0.241 3.200 0.942 0.313 50 0.311 4.492 0.903 0.048 55 0.482 7.529 0.768 0.339 61 0.520 8.668 0.729 0.595 OL 2 0.471 8.854 0.778 0.286 5 0.781 23.119 0.390 0.161 8 0.023 0.330 0.999 0.644 14 0.065 1.039 0.996 0.104 17 0.608 12.124 0.630 0.559 20 0.611 13.755 0.626 0.286 23 0.222 3.556 0.951 0.478 27 0.404 7.126 0.837 0.334 32 0.514 9.349 0.736 0.722 35 0.079 1.275 0.994 0.094 38 0.736 19.670 0.459 0.018 40 0.348 5.994 0.879 0.361 44 0.397 7.447 0.843 0.249 47 0.36 3 6.273 0.868 0.291 53 0.087 1.403 0.992 0.202 56 0.389 6.421 0.849 0.728 58 0.640 13.214 0.590 0.536 59 0.305 5.023 0.907 0.518 60 0.825 26.907 0.319 0.110 63 0.343 5.507 0.882 0.489 66 0.609 12.932 0.629 0.110 69 0.635 13.607 0.596 0.356

PAGE 112

112 Ta ble 3 14. Parameter estimates of the four factor model in Mongolia data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.216 4.696 0.890 0.045 4 0.555 12.224 0.690 0.913 7 0.035 0.641 0.999 0.982 13 0.439 9.97 0.806 0.38 6 16 0.540 12.405 0.711 0.460 19 0.474 11.447 0.777 0.014 22 0.332 7.176 0.889 0.390 26 0.557 14.509 0.695 0.151 33 0.309 6.945 0.796 0.060 36 0.377 8.628 0.864 0.265 39 0.450 10.83 0.799 0.306 42 0.490 11.169 0.758 0.718 46 0.429 9.999 0.819 0.342 49 0.404 8.843 0.837 0.611 52 0.255 5.615 0.616 0.806 57 0.624 14.857 0.788 0.099 62 0.449 10.555 0.797 0.085 65 0.207 4.363 0.959 0.248 67 0.264 5.735 0.932 0.061 PM 3 0.074 1.043 1.000 0.623 6 0.010 0.147 1.000 0.369 9 0.577 7.43 7 0.706 0.299 11 0.187 2.957 0.942 0.013 15 0.151 2.330 0.977 0.627 21 0.391 5.802 0.852 0.744 25 0.300 4.672 0.922 0.583 31 0.033 0.507 0.861 0.335 34 0.336 5.049 0.889 0.482 45 0.191 3.012 0.976 0.441 48 0.376 5.865 0.862 0.643 51 0.4 46 6.921 0.840 0.885 64 0.026 0.410 1.000 0.024 68 0.147 2.247 0.974 0.231

PAGE 113

113 Table 3 14. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.527 9.222 0.717 0.253 18 0.373 6.056 0.879 0.712 28 0.424 7.447 0.826 0.275 37 0.478 8.103 0.789 0.633 43 0.082 1.412 0.996 0.128 50 0.132 2.254 0.968 0.412 55 0.409 6.817 0.817 0.732 61 0.620 10.78 0.656 0.049 OL 2 0.341 7.038 0.877 0.120 5 0.412 8.724 0.824 0.130 8 0.121 2.331 0.988 0.314 14 0.367 6 .546 0.859 0.892 17 0.339 6.416 0.884 0.590 20 0.481 10.423 0.760 0.111 23 0.330 6.647 0.903 0.850 27 0.248 4.991 0.931 0.005 32 0.568 10.01 0.689 0.390 35 0.012 0.228 1.000 0.097 38 0.503 10.723 0.728 0.596 40 0.074 1.439 0.993 0.263 44 0.270 5.323 0.928 0.380 47 0.004 0.075 1.000 0.406 53 0.377 6.899 0.852 0.843 56 0.195 3.881 0.956 0.213 58 0.365 7.402 0.862 0.789 59 0.489 10.61 0.788 0.099 60 0.415 8.554 0.817 0.453 63 0.269 5.394 0.926 0.314 66 0.488 9.154 0.754 0.930 69 0.179 3 .361 0.971 0.664

PAGE 114

114 Table 3 15. Parameter estimates of the four factor model in Nigeria data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.293 3.616 0.914 0.631 4 0.394 4.896 0.844 0.726 7 0.197 2.554 0.961 0.697 13 0.436 6.314 0.810 0.352 16 0.428 6.152 0.817 0.045 19 0.436 5.932 0.810 0.280 22 0.377 4.995 0.858 0.375 26 0.542 7.476 0.706 0.213 33 0.346 4.621 0.881 0.259 36 0.272 3.653 0.926 0.055 39 0.412 5.899 0.830 0.003 42 0.205 2.689 0.958 0.223 46 0.463 6.587 0.786 0.492 49 0.342 4.642 0.883 0.245 52 0.374 4.311 0.86 0.827 57 0.497 6.956 0.753 0.417 62 0.272 3.585 0.926 0.115 65 0.248 3.324 0.939 0.534 67 0.336 4.671 0.887 0.670 PM 3 0.156 1.835 0.976 0.273 6 0.212 2.159 0.955 0.809 9 0.688 8.127 0.527 0.931 11 0.091 0.895 0.992 0.649 15 0.455 5.558 0.793 0.343 21 0.448 4.787 0.799 0.807 25 0.391 4.855 0.847 0.325 31 0.111 1.242 0.988 0.329 34 0.172 2.006 0.970 0.010 45 0.023 0.263 0.999 0.543 48 0.517 5.813 0 .732 0.888 51 0.385 4.505 0.852 0.707 64 0.380 4.451 0.855 0.401 68 0.098 1.121 0.990 0.507

PAGE 115

115 Table 3 15. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.467 4.865 0.781 0.096 18 0.613 5.976 0.625 0.681 28 0.309 3.260 0.904 0.090 37 0.067 0.678 0.995 0.391 43 0.182 1.684 0.967 0.584 50 0.453 4.347 0.794 0.505 55 0.149 1.464 0.978 0.571 61 0.306 3.315 0.906 0.026 OL 2 0.164 2.208 0.973 0.265 5 0.298 4.335 0.911 0.068 8 0.549 7.383 0.698 1. 124 14 0.282 3.625 0.920 0.713 17 0.891 12.381 0.206 1.563 20 0.647 12.109 0.582 0.483 23 0.274 3.342 0.925 1.034 27 0.259 3.539 0.933 0.259 32 0.589 8.417 0.653 1.175 35 0.307 3.528 0.906 1.138 38 0.582 8.473 0.661 0.921 40 0.018 0.238 1.000 0. 109 44 0.323 4.399 0.895 0.532 47 0.558 9.364 0.688 0.387 53 0.274 3.830 0.925 0.077 56 0.330 4.671 0.891 0.339 58 0.685 11.209 0.530 0.890 59 0.034 0.458 0.999 0.087 60 0.630 10.059 0.603 0.825 63 0.452 6.690 0.796 0.424 66 0.696 12.220 0.516 0 .732 69 0.426 5.874 0.818 0.774

PAGE 116

116 Table 3 16. Parameter estimates of the four factor model in Pakistan data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.345 5.057 0.881 0.055 4 0.324 4.541 0.895 0.217 7 0.442 6.16 7 0.805 0.523 13 0.358 5.129 0.872 0.188 16 0.412 6.04 0.830 0.377 19 0.153 1.969 0.976 0.437 22 0.226 3.071 0.949 0.063 26 0.416 5.988 0.827 0.205 33 0.291 4.035 0.915 0.253 36 0.184 2.387 0.966 0.289 39 0.246 3.308 0.940 0.102 42 0.451 6.91 4 0.797 0.163 46 0.378 5.268 0.857 0.383 49 0.247 3.385 0.939 0.363 52 0.400 5.787 0.840 0.102 57 0.429 6.479 0.816 0.336 62 0.230 3.118 0.947 0.238 65 0.157 2.124 0.975 0.146 67 0.225 3.091 0.949 0.022 PM 3 0.191 2.309 0.964 0.584 6 0.317 4.378 0.900 0.110 9 0.585 8.26 0.658 0.951 11 0.045 0.533 0.998 0.709 15 0.289 3.958 0.916 0.044 21 0.304 4.098 0.907 0.296 25 0.306 4.144 0.906 0.224 31 0.009 0.107 1.000 0.288 34 0.264 3.601 0.930 0.174 45 0.374 5.159 0.860 0.3 58 48 0.637 9.059 0.594 0.600 51 0.449 6.077 0.799 0.761 64 0.392 5.308 0.846 0.200 68 0.077 1.008 0.994 0.099

PAGE 117

117 Table 3 16. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.459 5.775 0.789 0.136 18 0.569 6 .903 0.676 0.294 28 0.407 5.113 0.834 0.005 37 0.337 4.169 0.886 0.118 43 0.026 0.307 0.999 0.393 50 0.111 1.303 0.988 0.187 55 0.228 2.758 0.948 0.429 61 0.655 7.884 0.571 0.166 OL 2 0.164 2.262 0.973 0.471 5 0.234 3.577 0.945 0.143 8 0.057 0.813 0.997 0.306 14 0.104 1.452 0.989 0.292 17 0.691 11.136 0.523 1.132 20 0.490 8.489 0.760 0.121 23 0.458 7.220 0.790 0.678 27 0.050 0.737 0.997 0.121 32 0.547 9.195 0.701 0.607 35 0.178 2.496 0.968 0.481 38 0.534 9.665 0.714 0.393 4 0 0.031 0.450 0.999 0.188 44 0.489 8.351 0.761 0.615 47 0.504 8.903 0.746 0.273 53 0.268 4.148 0.928 0.224 56 0.332 5.267 0.890 0.190 58 0.480 8.134 0.769 0.385 59 0.358 5.663 0.872 0.264 60 0.651 12.186 0.576 0.615 63 0.053 0.792 0.997 0.213 66 0.718 15.599 0.485 0.609 69 0.370 5.657 0.863 0.642

PAGE 118

118 Table 3 17. Parameter estimates of the four factor model in Philippines data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.042 0.482 0.998 0.737 4 0.418 5.7 77 0.826 0.431 7 0.211 2.453 0.956 0.704 13 0.300 3.868 0.910 0.452 16 0.461 6.704 0.788 0.515 19 0.233 2.984 0.946 0.142 22 0.279 3.573 0.922 0.336 26 0.550 8.471 0.698 0.079 33 0.426 5.607 0.819 0.218 36 0.307 3.758 0.906 0.641 39 0.500 7.2 60 0.750 0.193 42 0.442 5.478 0.805 0.721 46 0.638 10.195 0.593 0.248 49 0.488 5.925 0.762 0.573 52 0.188 2.316 0.965 0.618 57 0.518 6.510 0.732 0.867 62 0.379 5.195 0.857 0.218 65 0.139 1.743 0.981 0.110 67 0.269 3.594 0.928 0.290 PM 3 0.043 0.501 0.998 0.349 6 0.407 4.270 0.834 0.831 9 0.599 7.080 0.641 0.983 11 0.010 0.115 1.000 0.174 15 0.239 2.818 0.943 0.336 21 0.200 2.498 0.960 0.180 25 0.699 9.343 0.511 0.596 31 0.084 1.002 0.993 0.199 34 0.239 3.010 0.943 0.316 45 0.572 7.957 0.673 0.277 48 0.459 6.964 0.789 0.016 51 0.546 6.847 0.702 0.831 64 0.290 3.295 0.916 0.737 68 0.196 2.292 0.962 0.303

PAGE 119

119 Table 3 17. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.028 0.287 0.999 0.270 18 0.826 7.308 0.318 0.522 28 0.596 5.989 0.645 0.456 37 0.105 0.995 0.989 0.904 43 0.144 1.371 0.979 0.501 50 0.216 2.231 0.953 0.116 55 0.322 3.597 0.896 0.641 61 0.135 1.368 0.982 0.231 OL 2 0.105 1.476 0.989 0.079 5 0.77 0 15.564 0.408 0.696 8 0.311 4.617 0.903 0.091 14 0.228 2.857 0.948 0.787 17 0.593 7.406 0.648 1.238 20 0.469 7.598 0.780 0.206 23 0.112 1.147 0.987 1.421 27 0.569 9.721 0.676 0.257 32 0.503 5.781 0.747 1.173 35 0.527 7.250 0.722 1.068 38 0.636 11 .046 0.595 0.551 40 0.217 2.880 0.953 0.551 44 0.582 9.044 0.661 0.762 47 0.575 9.710 0.669 0.316 53 0.516 7.847 0.734 0.522 56 0.287 3.728 0.917 0.688 58 0.401 6.192 0.840 0.085 59 0.332 4.820 0.890 0.167 60 0.749 14.504 0.440 0.618 63 0.014 0. 180 1.000 0.438 66 0.558 8.077 0.688 0.923 69 0.198 2.802 0.961 0.120

PAGE 120

120 Table 3 18. Parameter estimates of the four factor model in Poland data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.528 9.317 0.721 0.081 4 0.498 7.970 0.752 0.366 7 0.277 4.040 0.841 0.546 13 0.268 3.861 0.928 0.511 16 0.760 15.592 0.422 0.219 19 0.226 3.153 0.949 0.219 22 0.488 7.817 0.761 0.273 26 0.698 13.914 0.513 0.055 33 0.500 8.089 0.750 0.049 36 0.129 1.813 0.983 0.189 3 9 0.444 7.074 0.803 0.316 42 0.248 3.378 0.938 0.526 46 0.610 11.374 0.628 0.151 49 0.484 7.869 0.766 0.251 52 0.340 5.131 0.884 0.046 57 0.285 3.905 0.919 0.612 62 0.025 0.353 0.999 0.113 65 0.297 4.230 0.912 0.664 67 0.384 5.856 0.853 0.287 PM 3 0.439 5.976 0.807 0.209 6 0.399 5.198 0.841 0.546 9 0.397 4.954 0.842 0.848 11 0.403 5.271 0.838 0.257 15 0.665 9.732 0.557 0.635 21 0.051 0.609 0.997 0.333 25 0.386 5.077 0.851 0.157 31 0.331 4.372 0.890 0.055 34 0.223 2.724 0.9 50 0.348 45 0.248 3.138 0.938 0.287 48 0.573 7.815 0.671 0.348 51 0.383 4.972 0.853 0.678 64 0.289 3.664 0.917 0.168 68 0.170 2.035 0.971 0.532

PAGE 121

121 Table 3 18. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.6 20 9.475 0.616 0.04 18 0.652 9.782 0.575 0.397 28 0.710 11.528 0.496 0.052 37 0.422 5.919 0.822 0.114 43 0.250 3.277 0.937 0.493 50 0.205 2.714 0.958 0.331 55 0.495 7.087 0.755 0.486 61 0.514 7.581 0.736 0.157 OL 2 0.377 5.820 0.858 0.215 5 0.487 7.213 0.763 0.566 8 0.384 6.138 0.853 0.245 14 0.545 8.571 0.704 0.776 17 0.655 14.110 0.571 0.435 20 0.397 6.142 0.842 0.605 23 0.255 3.790 0.935 0.339 27 0.287 4.362 0.918 0.58 32 0.589 11.238 0.653 0.576 35 0.573 10.529 0.672 0.46 38 0.501 8.866 0.749 0.151 40 0.400 6.692 0.840 0.263 44 0.327 5.137 0.893 0.017 47 0.480 7.893 0.770 0.385 53 0.277 4.299 0.923 0.069 56 0.226 3.398 0.949 0.339 58 0.544 9.032 0.704 0.664 59 0.020 0.286 1.000 0.573 60 0.712 15.952 0.492 0.081 63 0.022 0.301 1.000 0.865 66 0.647 13.677 0.582 0.116 69 0.492 8.480 0.758 0.287

PAGE 122

122 Table 3 19. Parameter estimates of the four factor model in Romania data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.063 0.7 89 0.996 0.433 4 0.578 8.891 0.666 0.687 7 0.125 1.578 0.984 0.341 13 0.200 2.545 0.960 0.223 16 0.536 8.523 0.712 0.164 19 0.446 6.151 0.801 0.569 22 0.228 2.925 0.948 0.617 26 0.631 10.610 0.601 0.256 33 0.429 5.688 0.816 0.207 36 0.442 6.1 14 0.805 0.361 39 0.366 4.989 0.866 0.077 42 0.365 4.671 0.866 0.541 46 0.618 9.173 0.618 0.003 49 0.539 7.853 0.710 0.445 52 0.192 2.430 0.963 0.371 57 0.515 7.226 0.735 0.825 62 0.571 8.577 0.674 0.323 65 0.116 1.486 0.987 0.051 67 0.196 2.553 0.961 0.035 PM 3 0.439 5.548 0.807 0.293 6 0.408 4.476 0.833 0.623 9 0.322 3.564 0.897 0.767 11 0.016 0.183 1.000 0.758 15 0.494 6.685 0.756 0.094 21 0.064 0.773 0.996 0.190 25 0.522 6.633 0.728 0.426 31 0.205 2.456 0.958 0.296 3 4 0.324 4.150 0.895 0.161 45 0.388 5.041 0.850 0.010 48 0.575 7.063 0.670 0.438 51 0.402 4.372 0.838 1.097 64 0.206 2.481 0.958 0.304 68 0.093 1.052 0.991 0.629

PAGE 123

123 Table 3 19. Continued Trait/Item Factor loading Estimate/SE Residual Variance Th reshold TF 12 0.198 1.926 0.902 0.240 18 0.313 3.139 0.672 0.019 28 0.573 5.457 0.958 0.296 37 0.377 3.806 0.858 0.478 43 0.508 4.846 0.741 0.336 50 0.168 1.574 0.972 0.183 55 0.382 3.545 0.854 0.795 61 0.427 4.135 0.817 0.119 OL 2 0.412 6 .433 0.830 0.296 5 0.616 11.152 0.621 0.155 8 0.374 5.443 0.860 0.491 14 0.455 6.978 0.793 0.662 17 0.750 15.052 0.438 0.770 20 0.678 13.051 0.540 0.019 23 0.556 8.704 0.691 0.771 27 0.515 8.905 0.734 0.410 32 0.824 18.742 0.320 0.745 35 0.476 7 .444 0.773 0.594 38 0.611 11.582 0.626 0.371 40 0.091 1.281 0.992 0.016 44 0.421 6.681 0.823 0.097 47 0.423 6.562 0.821 0.022 53 0.147 2.093 0.978 0.151 56 0.390 5.556 0.848 0.755 58 0.604 11.197 0.636 0.251 59 0.011 0.157 1.000 0.273 60 0.634 11.598 0.598 0.257 63 0.042 0.579 0.998 0.454 66 0.655 11.924 0.571 0.524 69 0.361 5.379 0.869 0.371

PAGE 124

124 Table 3 20. Parameter estimates of the four factor model in Samoa data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.016 0.185 0.999 1.200 4 0.365 6.041 0.867 0.468 7 0.709 13.34 0.497 0.157 13 0.590 9.023 0.652 0.385 16 0.020 0.276 0.999 0.063 19 0.510 9.732 0.740 1.326 22 0.286 4.159 0.918 0.496 26 0.214 2.878 0.954 0.517 33 0.437 8.393 0.809 0.6 82 36 0.414 6.735 0.828 0.496 39 0.321 4.085 0.897 0.598 42 0.527 10.369 0.722 0.824 46 0.540 9.785 0.708 0.372 49 0.447 8.268 0.801 0.454 52 0.435 7.275 0.811 0.824 57 0.245 3.51 0.940 0.517 62 0.500 10.105 0.750 0.636 65 0.363 6.556 0.868 0.789 67 0.176 2.581 0.969 0.038 PM 3 0.217 3.179 0.953 0.228 6 0.201 2.923 0.960 0.824 9 0.388 6.486 0.849 1.115 11 0.069 1.016 0.995 0.419 15 0.017 0.232 0.999 0.161 21 0.398 6.497 0.842 0.598 25 0.332 4.728 0.890 1.015 31 0.369 5. 851 0.864 0.524 34 0.232 2.969 0.946 0.247 45 0.354 4.584 0.875 1.282 48 0.519 7.989 0.731 0.798 51 0.579 10.314 0.665 0.954 64 0.011 0.142 0.999 0.575 68 0.311 4.407 0.903 0.392

PAGE 125

125 Table 3 20. Continued Trait/Item Factor loading Estimat e/SE Residual Variance Threshold TF 12 0.209 2.92 0.956 0.138 18 0.284 4.243 0.920 0.461 28 0.231 3.21 0.947 0.132 37 0.410 4.666 0.832 1.555 43 0.264 3.915 0.930 0.399 50 0.450 7.424 0.797 0.842 55 0.456 8.32 0.792 0.674 61 0.630 9.486 0.603 0.126 OL 2 0.282 3.733 0.921 0.496 5 0.432 4.323 0.813 0.944 8 0.276 4.842 0.924 0.706 14 0.226 2.468 0.949 1.457 17 0.492 8.005 0.758 1.576 20 0.193 2.981 0.963 0.273 23 0.527 8.523 0.722 1.695 27 0.604 10.98 0.636 0.069 32 0.494 6.787 0 .756 1.812 35 0.279 3.631 0.922 0.994 38 0.304 3.471 0.907 1.282 40 0.432 8.181 0.814 0.379 44 0.255 3.422 0.935 0.352 47 0.216 3.206 0.953 0.561 53 0.490 8.739 0.760 0.781 56 0.381 5.04 0.855 1.240 58 0.370 5.805 0.863 0.915 59 0.494 9.413 0.75 6 1.058 60 0.463 7.534 0.786 1.047 63 0.363 6.221 0.868 0.365 66 0.499 10.2 0.751 1.069 69 0.231 3.534 0.947 0.532

PAGE 126

126 Table 3 21. Parameter estimates of the four factor model in Singapore data Trait/Item Factor loading Estimate/SE Residual Varia nce Threshold EI 1 0.272 4.427 0.926 0.191 4 0.641 12.564 0.589 0.644 7 0.377 6.184 0.858 0.378 13 0.508 9.000 0.742 0.446 16 0.661 13.216 0.563 0.378 19 0.417 7.120 0.826 0.408 22 0.508 9.815 0.742 0.023 26 0.680 14.776 0.538 0.397 33 0. 164 2.592 0.973 0.175 36 0.377 6.444 0.858 0.062 39 0.487 9.172 0.763 0.271 42 0.566 10.092 0.680 0.686 46 0.497 9.016 0.753 0.352 49 0.495 9.245 0.755 0.235 52 0.309 5.061 0.905 0.283 57 0.620 13.278 0.616 0.288 62 0.434 7.618 0.812 0.194 6 5 0.189 2.932 0.964 0.339 67 0.403 7.158 0.838 0.199 PM 3 0.408 5.620 0.833 0.013 6 0.325 4.294 0.894 0.317 9 0.494 6.482 0.756 1.014 11 0.386 5.176 0.851 0.086 15 0.378 5.185 0.857 0.120 21 0.186 2.416 0.965 0.016 25 0.428 5.911 0.817 0 .185 31 0.338 4.523 0.886 0.003 34 0.188 2.411 0.965 0.271 45 0.182 2.292 0.967 0.287 48 0.272 3.569 0.926 0.276 51 0.627 8.740 0.606 0.753 64 0.240 3.085 0.942 0.086 68 0.230 3.012 0.947 0.062

PAGE 127

127 Table 3 21. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.567 8.595 0.678 0.068 18 0.469 6.324 0.780 0.609 28 0.640 9.976 0.590 0.060 37 0.460 6.309 0.788 0.578 43 0.073 0.846 0.995 0.746 50 0.238 3.185 0.944 0.096 55 0.446 5.195 0.801 1.004 61 0.593 9.024 0.649 0.279 OL 2 0.371 5.787 0.862 0.534 5 0.596 12.167 0.645 0.419 8 0.324 4.680 0.895 0.699 14 0.383 3.934 0.854 1.456 17 0.623 8.206 0.612 1.357 20 0.646 13.530 0.583 0.425 23 0.339 4.805 0.885 1.005 27 0.356 5.659 0.874 0.049 32 0. 671 10.358 0.550 1.250 35 0.406 4.718 0.836 1.309 38 0.685 13.941 0.530 0.753 40 0.164 2.403 0.973 0.454 44 0.403 6.207 0.838 0.641 47 0.356 5.958 0.873 0.112 53 0.333 5.323 0.889 0.013 56 0. 300 4.636 0.910 0.435 58 0.502 8.476 0.748 0.510 59 0.2 26 3.432 0.949 0.068 60 0.823 19.564 0.323 0.827 63 0.214 3.275 0.954 0.089 66 0.689 12.767 0.525 0.845 69 0.540 9.277 0.708 0.691

PAGE 128

128 Table 3 22. Par ameter estimates of the four factor model in U.S. data Trait/Item Factor loading Estimate/SE Res idual Variance Threshold EI 1 0.428 7.525 0.817 0.217 4 0.575 7.966 0.670 1.146 7 0.207 3.159 0.957 0.524 13 0.362 5.665 0.869 0.490 16 0.619 13.517 0.617 0.111 19 0.274 4.485 0.925 0.015 22 0.476 8.681 0.773 0.233 26 0.666 15.751 0.556 0.2 64 33 0.296 4.669 0.912 0.496 36 0.258 4.041 0.934 0.429 39 0.241 3.821 0.942 0.105 42 0.377 6.220 0.858 0.396 46 0.627 13.648 0.607 0.337 49 0.488 9.156 0.762 0.238 52 0.380 6.342 0.856 0.311 57 0.569 10.346 0.677 0.706 62 0.215 3.447 0.954 0.166 65 0.332 5.503 0.890 0.327 67 0.405 7.196 0.836 0.187 PM 3 0.318 4.471 0.899 0.305 6 0.440 6.498 0.806 0.300 9 0.344 4.804 0.881 0.524 11 0.008 0.111 1.000 0.565 15 0.550 8.381 0.697 0.111 21 0.539 8.113 0.709 0.116 25 0.430 6.287 0 .815 0.050 31 0.153 2.111 0.977 0.030 34 0.455 6.745 0.793 0.181 45 0.358 5.026 0.872 0.300 48 0.496 7.423 0.754 0.327 51 0.473 6.829 0.776 0.479 64 0.449 6.655 0.798 0.111 68 0.087 1.185 0.993 0.402

PAGE 129

129 Table 3 22. Continued Trait/Item Factor lo ading Estimate/SE Residual Variance Threshold TF 12 0.535 8.502 0.714 0.187 18 0.677 11.983 0.542 0.269 28 0.674 11.732 0.546 0.005 37 0.471 7.331 0.778 0.342 43 0.166 2.245 0.972 0.429 50 0.384 5.873 0.853 0.065 55 0.424 6.350 0.820 0.674 61 0.696 12.552 0.516 0.010 OL 2 0.540 10.764 0.709 0.161 5 0.463 8.560 0.786 0.085 8 0.414 6.831 0.829 0.650 14 0.353 5.453 0.875 0.745 17 0.643 12.977 0.587 0.650 20 0.603 12.760 0.636 0.166 23 0.362 5.431 0.869 0.863 27 0.317 5.365 0.899 0.0 40 32 0.659 13.599 0.566 0.559 35 0.418 6.445 0.825 0.856 38 0.606 12.432 0.633 0.202 40 0.327 5.326 0.893 0.380 44 0.397 6.833 0.842 0.290 47 0.329 5.368 0.892 0.440 53 0.292 4.799 0.915 0.085 56 0.359 6.162 0.871 0.342 58 0.573 11.894 0.671 0. 060 59 0.328 5.591 0.892 0.025 60 0.765 18.592 0.414 0.402 63 0.324 5.531 0.895 0.070 66 0.534 10.532 0.715 0.161 69 0.408 6.975 0.834 0.451

PAGE 130

130 Table 3 23. Parameter estimates of the modified four factor model in Venezuela data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.317 4.752 0.900 0.156 4 0.558 7.413 0.689 0.997 7 0.072 0.865 0.995 0.787 13 0.313 4.614 0.902 0.132 16 0.465 7.231 0.784 0.444 19 0.287 4.112 0.917 0.129 22 0.271 3.803 0.927 0.037 2 6 0.518 8.377 0.731 0.250 33 0.421 6.532 0.823 0.380 36 0.438 5.154 0.808 1.022 39 0.388 5.233 0.850 0.733 42 0.512 7.147 0.738 0.861 46 0.365 5.555 0.867 0.138 49 0.314 4.497 0.902 0.332 52 0.427 6.098 0.818 0.525 57 0.827 15.439 0.315 0.93 1 62 0.482 7.459 0.768 0.402 65 0.106 1.441 0.989 0.408 67 0.418 6.552 0.825 0.260 PM 3 0.437 5.735 0.809 0.285 6 0.399 4.189 0.841 1.011 9 0.033 0.359 0.999 0.674 11 0.237 2.748 0.891 0.268 15 0.134 1.607 0.982 0.339 21 0.275 3.252 0.9 25 0.623 25 0.208 2.545 0.957 0.288 31 0.249 3.234 0.938 0.034 34 0.182 2.256 0.967 0.073 45 0.091 1.133 0.992 0.028 48 0.315 3.973 0.901 0.272 51 0.153 1.787 0.976 0.625 64 0.087 1.069 0.992 0.289 68 0.152 1.863 0.977 0.971

PAGE 131

131 Table 3 23. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.330 3.735 0.891 0.268 18 0.416 4.779 0.827 0.555 28 0.679 7.01 0.539 0.003 37 0.108 0.856 0.988 1.370 43 0.117 1.293 0.986 0.203 50 0.151 1.613 0.977 0.229 55 0.483 4.611 0.767 0.676 61 0.487 5.527 0.763 0.125 OL 2 0.300 4.691 0.910 0.119 5 0.857 25.896 0.266 0.485 8 0.357 5.377 0.872 0.587 14 0.090 1.074 0.992 0.981 17 0.680 13.419 0.538 0.951 20 0.589 11.302 0.654 0.018 23 0.270 3.949 0.927 0.680 27 0.293 4.415 0.910 0.442 32 0.704 14.269 0.504 0.895 35 0.142 1.89 0.980 0.921 38 0.426 6.656 0.818 0.260 40 0.172 2.501 0.971 0.307 44 0.441 7.545 0.806 0.067 47 0.591 11.344 0.651 0.162 53 0.405 6.503 0.836 0.340 56 0.424 6.772 0.821 0.49 0 58 0.481 8.038 0.768 0.667 59 0.259 3.801 0.933 0.451 60 0.818 22.63 0.332 0.510 63 0.425 7.142 0.819 0.179 66 0.674 13.225 0.545 0.743 69 0.572 10.729 0.672 0.490

PAGE 132

132 Table 3 24. Parameter estimates of the modified four factor model in Zimbabw e data Trait/Item Factor loading Estimate/SE Residual Variance Threshold EI 1 0.100 1.284 0.990 0.296 4 0.418 5.057 0.825 0.782 7 0.264 3.228 0.930 0.670 13 0.143 1.818 0.979 0.416 16 0.104 1.358 0.989 0.028 19 0.151 1.949 0.977 0.323 22 0.114 1.472 0.987 0.067 26 0.162 2.161 0.974 0.033 33 0.029 0.363 0.999 0.342 36 0.288 3.617 0.917 0.406 39 0.285 3.783 0.919 0.154 42 0.323 4.164 0.896 0.412 46 0.047 0.593 0.998 0.269 49 0.163 2.146 0.974 0.157 52 0.217 2.774 0.953 0.191 5 7 0.612 8.838 0.625 0.480 62 0.562 7.875 0.684 0.482 65 0.001 0.010 0.999 0.202 67 0.219 2.553 0.952 0.746 PM 3 0.226 3.200 0.949 0.294 6 0.243 2.946 0.941 0.813 9 0.343 4.562 0.882 0.710 11 0.270 3.288 0.927 0.783 15 0.084 1.147 0.99 3 0.152 21 0.114 1.564 0.987 0.236 25 0.278 4.005 0.923 0.246 31 0.502 6.967 0.748 0.727 34 0.231 3.267 0.947 0.036 45 0.300 4.247 0.910 0.416 48 0.421 6.211 0.823 0.519 51 0.340 4.609 0.885 0.501 64 0.300 4.047 0.916 0.235 68 0.423 6.129 0.821 0.431

PAGE 133

133 Table 3 24. Continued Trait/Item Factor loading Estimate/SE Residual Variance Threshold TF 12 0.194 1.897 0.962 0.227 18 0.431 4.284 0.814 0.495 28 0.390 3.832 0.848 0.350 37 0.221 2.024 0.951 0.505 43 0.116 1.148 0.987 0.196 50 0.218 2.117 0.953 0.350 55 0.409 3.759 0.833 0.584 61 0.003 0.034 0.999 0.074 OL 2 0.010 0.149 0.999 0.275 5 0.123 1.832 0.985 0.183 8 0.380 4.619 0.856 1.121 14 0.300 3.456 0.915 0.910 17 0.687 8.529 0.527 1.579 20 0.485 7.235 0.765 0 .811 23 0.694 11.235 0.518 1.015 27 0.324 4.908 0.895 0.461 32 0.450 6.359 0.797 1.031 35 0.493 7.234 0.757 0.922 38 0.641 10.835 0.589 0.989 40 0.128 1.891 0.984 0.204 44 0.563 9.773 0.683 0.768 47 0.388 6.254 0.849 0.472 53 0.311 4.875 0.904 0.3 33 56 0.062 0.936 0.996 0.232 58 0.512 8.237 0.738 0.826 59 0.054 0.807 0.997 0.113 60 0.594 9.981 0.647 0.925 63 0.443 7.407 0.803 0.425 66 0.585 9.683 0.657 0.916 69 0.370 5.333 0.863 0.790

PAGE 134

134 CHAPTER 4 DISCUSSION The purpose of this study was to examine cross nationally the four factor structure of temperament as measured by the SSQ -a somewhat popular measure of 19 96 results psychometr ic properties have been supported largely by U.S. data Although SSQ data o n children from other countries are available, there have been few attempts to examine the ir factor structure. The validity of SSQ is grounded on both theoretical and empirical evidences. The develop ment of SSQ relies on a well established psychological type theory that was introduced by Jung (1971/1921) and extended by Briggs and Myers (MBTI: 1976). This theory defined the framework of SSQ as a measure of four bipolar temperament traits. internal structure, the four factor model of temperament has been consistently supported through exploratory and confirmatory factor analyses (Callueng, de Carvalho, Isobe, & Oakland, 2012; Benson, Oakland, & Shermis, 2009; Strafford & Oakland, 1996 ; Oakl and, Glutting, & Horton, 1996). Validity of the SSQ is further supported by other sources. At the item level, response patterns generally were similar among Hispanic, African American, and White students ( Stafford, 1994). External validation procedures ind icated that temperament preferences may influence career choices, class preferences, and involvement in school activities ( Oakland, Glutting, & Horton, 1996 ). Temperament preferences are associated to personal values and similar traits measured by the MBTI Temperament preferences are independent of cognitive traits, including achievement and intelligence ( Oakland, Glutting, & Horton, 1996).

PAGE 135

135 Oakland, & Shermis, 2009) and Rasch me asurement modeling ( Mpofu, Oakland, & Gwirayi, 2010 ) that reported stability of the four f actor model. The current study attempted to extend the validity evidence of the four factor model by applying factor analytic procedures at the item level. Additional ly, t his study is a large scale cross national investigation that includes data from 17,867 children in 21 countries that represent major world regions. This large data has implications for the use of chi square ( 2 ) statistics. Overall model fit of the data from each country was assessed using root mean square error approximation (RMSEA) as primary index of fit, and comparative fit index (CFI) and Tucker Lewis index (TLI) as secondary indices of fit. Although 2 values also are reported, they are not considered to be a primary fit criterion because they tend to be less useful with larger sample size (Hopwood & Bonnelan, 2010) General findings from confirmatory factor analyse s (CFA) using data from each country, revealed RMSEA values that were thus suggesting a good fit of the four factor model of temperament. However, all CFI and TLI indices failed to meet the cut .90) and thus suggesting a poor fit to the model in both the initial and mod ified solutions. As expected, the 2 values did not confirm the hypothesized factor model Taken as a whole, the overall fit indices indicate that the data from the 21 countries did not fit the four factor model of temperament reasonably well. Consequently a test of invariance was not implemented. The inadequate fit of the four factor model of temperament in the 21 countries is not consistent with previous findings (Callueng, de Carvalho, Isobe, & Oakland, 2012;

PAGE 136

136 Benson, Oakland, & Shermis, 2009; Strafford & Oakland, 1996 ; Oakland, Glutting, & Horton, 1996) They reported a good fit of the four factor model in da ta from children in the U.S. as well as children in other countries The discrepancy between the findings of the current study and the earlier studi es may be attributed to the use of different as indicators of extroverted introverted (EI), practical imaginative (PM), thinking feeling (TF), and organized flexible (O L) traits by combining single dichotomous items into parcels in order to produce greater variance ( Brown, 2006; Zwick, 1987). The use of item parceling s upport s the four factor model of temperament Moreover, the bipolar traits were non overlapping in data from children in the U.S. (Oakland, Glutting, & Horton, 1996; Strafford & Oakland, 1996) and children from other countries ( Callueng, de Carvalho, Isobe, & Oakland, 2012; Benson, Oakland, & Shermis, 2009). In contrast the current study employed item lev el factor analysis for categorical var iable methodology (CVM). Some believe this analysis may be more appropriate for dichotomous test items (Bandalos, 2008; Ivanova et. al., 2007; Yu & Muthen, 2002) However, factor analysis for CVM is robust and stable when dealing with a maximum of 25 30 variables (Muthen & Kaplan, 1992). The 63 SSQ items constitute 63 non overlapping variables used in the CFA. Thus, this figure exceeds the limits set for the use of a factor analysis with CVM. This large number of varia bles used as indicators of a relatively small number of latent factors has been unable to meet the requirements for model fit with other data sets. Thus, this issue is not unique to SSQ data. For example, similar findings are reported in a study that exami ned the factor structure of the Myers Briggs Type Indicator, also a measure of four bipolar temperaments albeit in

PAGE 137

137 adults (Bess, Harvey, & Swartz, 2003). Findings and the results from simulation work on CVM (Flora & Curran, 2004) when used with robust app roaches to evaluate fit indices of models with categorical an d non normal manifest variables suggest item parceling may be an acceptable alternative strategy (Brown, 2006). Item parceling seemingly works well with SSQ data and may be more suitable than CVM that is prone to model misspecification s when used with large number of items for a simple factor structure. Moreover, there is little support for the accuracy and stability of fit indices for f actor solution estimates when using the mean and variance ad justed weighted least squares estimator (WLSMV) (Bauducel & Herzberg, 2009). More specifically, use of CFI and TLI as overall fit indices in CFA for categorical data is found to be less stable and robust compared to RMSEA (Yu and Muthen, 2002) Following t his cautionary note, the current study and other studies have considered CFI and TLI as secondary fit indices to RMSEA (e.g., Ivanova et. al., 2007 ) Aside from overall fit indices, item factor loadings impact model fit. Item reliability in turn, influenc es factor loadings (Nunnally & Berstein 1994; Gorsuch, 1983). The f indings of the current study indicate considerable v ariability i n the number of items that have appreciable loadings (i.e., 30 ) among the countries, ranging from 31 (Zimbabwe) to 54 (U.S.), with an average of 43 item (68%) indicators distributed across the four temperament traits. Items that measure EI and OL traits have higher factor loadings than those that measure PM and TF traits. F actor loadings of items measurin g EI and OL have been enhanced by their higher reliability coefficients. C onversely, lower

PAGE 138

138 factor loadings for items that measur e PM and TF have been attenuated by their lower reliability coefficients. Factor i ntercorrelations provide important data to assist us when evaluating the fit of the four factor model of temperament. Previous studies reported that the bipolar traits measured in the SSQ are not overlapping and thus provide evidence that they are somewhat distinct constructs (Benson, Oakland, & Shermis, 2009; Strafford & Oakland, 1996 ; Oakland, Glutting, & Horton, 1996). In contrast, the current study found mixed results on whether the bipolar traits overlap. In general, data from 15 (71%) of the 21 countr ies displayed independence of the bipolar traits and the remaining 6 (29%) countries (i.e., China, Egypt, Gaza, Hungary, Iran, & Samoa) displayed multicollinearity and thus overlap on what the bipolar traits measure. Furthermore, multicollinearity in China Egypt, and Gaza was severe, resulting to a non positive correlation matrix or an improper solution. Notably, PM and OL are associated more frequently in countries with traditions and value orientations that are consistent with a collectivist culture. Nig eria, Pakistan, Philippines, Romania, Samoa, Venezuela, and Zimbabwe are known for their collectivistic orientations. This finding concurs with the belief that culture may influence The non positive solution displayed in data from Egypt and Gaza may be associated with turbulent situations in these countries at the time the data were gathered as well as test adaptation factors since the SSQ in both countries used the Arabic language.

PAGE 139

139 The non positive solution displayed in data from C hina may be associated with some cultural qualities common to Chinese people such as filial piety, harmony, Ren Qing (relationship orientation), modernization, thrift vs. extravagance, Ah Q Mentality (defensiveness), and face (Cheung et al., 1996). A failu re to find fit of the four factor model for Chinese children as well as for the big five personality factor model for Chinese adults may be due to similar qualities (Cheung et al., 2001). Culture may shape temperament and personality qualities in Chinese p eople. Lastly, the model misspecifications in the CFA solutions may have been influenced by sample size s The use of as ymptotic distribution free estimator (e.g., the robust weighted least square estimator, WLS ) requires an unbiased and large sample size s (Byrne, 2012). Moreover when using WLSMV guidelines for sample size requirement s have not been fully established (Brown, 2006) R ule s of thumb for minimum to moderate (i.e. 5 to 10 participants per indicator) sample size s may not be su fficient for CFA when using robust WLS estimator (Byrne, 2012). Future studies that examine the factor structure of the SSQ at the individual item level can use power analysis as a viable method to determine the appropriate and adequate sample size s Power analysis is st atistically defined as 1 minus the probability to commit a Type 2 error, with adequate power set at .80 (Cohen, 1988). Determining a sample size based on adequate power will help ensure precise parameter estimates and generalizability of model specificatio ns (Schmitt, 2011).

PAGE 140

140 Implications International Research Interest in multicultural issues has spawned research and testing practices within and between countries (Byrne et. al., 2009). Temperament data utilized in the current study were collected through collaborative efforts of researchers in 21 countries. As a non their development and gender differences. Scholars and practitioners in some countries (e.g., Kuwait, Republic of China, Romania, and South Korea) that have adapted and normed the SSQ use it in research and school based practices, including classroom learning, instructional design, and career exp loration. The current study is a product of well established international collaboration that was forged through common interest, scholarship, and professional commitment. Cross national studies are important when attempting to validate theoretical models. The five factor model of personality (McCrae & Costa, 2000; McCrae & Costa, 1977) and the 8 syndrome taxonomic model for youth psychopathology (Ivanova et. al., 2007) are based on theoretical models that successfully transcended linguistic and cultural di fferences. In a similar vein, the assessment of temperament as examined in the current study was able to transcend differences in language, religion, ethnicity, and sociopolitical orientation. The use of a common model of temperament facilitates communicat ion and collaboration for educational and mental health professionals from different countries.

PAGE 141

141 Structural and Measurement Equivalence Structural and measurement equivalence constitute s a primary strand in cross national research. An array of approaches h as been introduced to establish different but complimentary forms of equivalence: construct, functional, linguistic, and metric (Poortinga, 1995; van de Vijver & Tanzer, 1997; van de Vijver & Leung, 2000 ) Construct equivalence is achieved when there is hi gh degree of factorial structure invariance across groups (Schmitt, Allik, McCrae, & Benet Martinez, 2007). Functional invariance is attained when factors or scales are associated with other variables in similar ways (Schmitt, Allik, McCrae, & Benet Martin ez, 2007). Linguistic equivalence is demonstrated when the scales or dimensions of a test are internally consistent across languages (Caprara, Barbaranelli, Bermudez, Maslach, & Ruch, 2000). Metric equivalence is achieved when items administered in differe nt lang uages function in the same way (Ramirez Esparza, Gosling, Benet Mart inez, Porter, & Pennebaker, 2006 ). factor model of temperament has been assessed through exploratory factor analysis (Oakland, Glutting, & Horton, 1996), differentia l item functioning ( Strafford & Oakland, 1996 ), Rasch measurement modeling ( Mpofu, Oakland, & Gwirayi, 2010), and confirmatory factor analysis using item parcels (Benson, Oakland, & Shermis, 2009). These analytical strategies confirmed the stability of tem perament as constituting four independent bipolar constructs: extraverted introverted, practical imaginative, thinking feeling, and organized flexible. The current study examined the factor structure of temperament through the use of factor analysis that f ocused on dichotomous items. Because of the restricted variance inherent in the use of dichotomous items, the mean and variance adjusted weighted least sq uares (WLSMV) estimation was used along with tetrachoric correlations as input

PAGE 142

142 to confirmatory factor analysis. This analytic approach is robust in correcting observed covariances in binary data. Current methods to assess construct validity of tests with dichotomous items have established the relationship between factor analysis and item response theory ( IRT) (Kamata & Bauer, 2008). As shown in the above results, the limited variance inherent in the dichotomous item format of the SSQ may not be viable for the requirements of the CFA cture may be examined using IRT an a pproach that is more suitable for dichotomous items (Wirth & Edwards, 2007) such as those found on the SSQ. IRT implements a two parameter logistic model as a likelihood based estimation model for logistic approximation to the normal ogive (Wirth & Edwards 2007). The use of threshold as a difficulty index provides important statistical information of the latent response model that takes account of the dichotomous nature of the items (Kamata & Bauer, 2008). When applied to SSQ data, preference for temperame nt quality (i.e., EI, PM, TF, or OL) is conceptualized as a function of the probability of endorsing an item response that is plotted in an item characteristic curve or trace line. Test Validity Tes t validity, an essential test characteristic, is reflected in its theoretical and empirical support (AERA, APA, & NCME, 1999). Evidences of validity can be obtained through various sources that, in their combined use, can complement each other to The i ncreasing popularity of SSQ in other countries as evidenced in the current study indicates interest in the possible usefulness of temperament data in promoting

PAGE 143

143 self understanding, counseling, assessing learning styles, instructional matching, exploring voc ational needs, and facilitating research and program evaluation. In addition to factor structure, t he validity of the SSQ adapted versions can be enhanced by establishing a network of variables that are associated with temperament (i.e., convergent validit y) and the variables that are not associated with temperament (i.e., discriminant validity). Validity evidence of the SSQ adapted versions also can be demonstrated through group comparisons (i.e., gender, age, clinical diagnoses, etc.). Limitations Severa l limitations of this study need to be acknowledged First, sample sizes varied, ranging from 7,902 for the U.S. to 253 for Israel. DiStefano (2002) suggested using large sample size s when using the w eighted least squares estimator for CFA to ensure accura cy and generalizability of results In addition, larger sample size s increase statistical power and precision of parameter estimates in CFA (Schmitt, 2011). Hence, results from countries with relatively small sample size s (e.g., Israel ) may produce biased estimates and are not considered conclusive. Second, samples in all countries except that of the U.S. were drawn from a specific geographical area and thus do not represent the general population of children in those countries. Whether responses of these are fully representative of children nationally cannot be verified Third, the SSQ English version was used to collect temperament da ta in children from Nigeria, Pakistan and Singapore Although children in these countries attended English speaking schoo ls, it is not known whether responses in the English language and in a native language differ for bilingual children.

PAGE 144

144 Fourth, only the four factor model of temperament was tested in all countries and no competing model was introduced. Other measurement mo dels that have theoretical and substantive support should have been used to determine the best fitting model in as much as the four factor model of temperament in all the countries did not provide a good fit. Fifth, a modified solution was performed for th e countries to address misspecifications of the CFA model. Modifications in a CFA model are based on theoretical and empirical evidence. Since no previous studies were available to guide model modifications, content meaning at the item level provided the b asis on which to carry out a modified solution Conclusions Despite these limitations, results from this study contribute to the literature on the factorial validity of the SSQ. Although item level CFA did not support the cross national generalization of the four factor structure of temperament, the results can be considered format restricts variance that can affect observed covariances and produce biased CFA estimates. Moreover, the relatively small sample sizes in most cou ntries may not have met the large sample size requirement s when using the WLSMV estimator. Thus, attempts to find a model fit using CFA can be problematic. In previous studies that used item parcels i n factor analysis, differen tial item functioning, and IRT provided theoretical support for the four bipolar temperament model. Hence, we can conclude that the lack of support for the factorial structure of the SSQ as reported in this study is tentative and future studies can

PAGE 145

145 Future studies also may focus on collecting additional empirical evidences on the with childr en from the U.S. that examine whether there is support on the internal structure and functional validity of the SSQ. using differential item functioning, item response theor y, and factor analysis using item parcels. Utilizing these other methods with data from other countries may find support for the construct validity of the SSQ and add to existing literature. The four temperament traits are known to function independently. Yet the findings from this study indicate these traits highly overlap in data from China, Egypt, and Gaza Subsequent research can examine the factor structure of SSQ in these countries using larger sample size s To overcome restricted variance in dichot omous items, SSQ item format could be modified into a continuous scale to eliminate its dichotomous nature. Factor structure of the SSQ with this continuous scale format can be examined to determine if the four factor model of temperament fits well with th is item format.

PAGE 146

146 LIST OF REFERENCES American Psychological Association, American Educational Research Association, & National Council on Measurement in Education (1999). Standards for educational and psychological testing Washington, DC: American Educati onal Research Association. Bandalos, D. L. (2008). Is parceling really necessary? A comparison or results from item parceling and categorical variable methodology. Structural Equation Modeling 15, 211 240. Bandalos, D. L., & Finney, S. J. (2001). Item par celing issues in structural equation modeling. In G. A. Marcoulides & R. E. Schumacker (Eds.), New developments and techniques in structural equation modeling (pp. 269 296). Mahwah, NJ: Lawrence Erlbaum Associates Banducel, A. & Herzberg, P. Y. (2 009). On the performance of maximum likelihood versus means and variance adjusted weighted least squares estimation in CFA. Structural Equation Modeling: A Multidisciplinary Jo urnal, 13 (2), 186 203. Barkley, R. A. (1997). manual for assessment and parent training New York: Guilford Press. Bassett, K. & Oakland, T. (2009). Temperament preferences for c hildren ages 8 through 17 in a nationally represented sample. In J. Kaufman (Ed ). I n telligent testing: Integrating psychol ogical theory and clinical practice (pp 30 52). Boston: Cambridge University Press Beckner, P.A. (1990). The learning styles of achieving and underachieving gifted and talented tenth grade students Univers ity, Sacramento, CA. Benson, N., Oakland, T., & Shermis, N. (2009) Cross temperament. Journal of Psychoeducational Assessment 27, 3 16. Berry, J.W. Poortinga, Y.H., Segall, M.H., & Dasen, P.R. (1992 ). Cross cultural psych ology: Research and applications Cambridge: Cambridge University Press. Bess, T. L., & Harvey, R. J. (2002). Bimodal score distributions and the Myers Briggs Type Indicator: Fact or artifact? Journal of Personality Assessment 78, 176 186. Bireley, M. (19 91). Learning styles: One way to help gifted adolescents understand and choose life styles. In M. Bireley & J. Genshaft (Eds.), Understanding the gifted adolescent: Educational, developmental, and multicultural issues (pp. 189 200). New York: Teacher Colla ge Press.

PAGE 147

147 Briggs, K. C. & Myers, I. B. (1976). Myers Briggs Type Indicator: Form F Palo Alto, CA: Consulting Psychologists Press. Brown, T. A. (2006). Confirmatory factor analysis for applied research New York: Guilford. Brown, M. W. & Cudeck, R. (1993) Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136 162). Newbury Park, CA: Sage. Buss, A. H. (2011). Pathways to individuality: Evolution and development of personality traits. Washin gton, DC: APA. Byrne, B. M. (1994). Structural equation modeling with EQS and EQS Windows: Basic concepts, applic ations, and programming. Thousand Oaks, CA: Sage Byrne, B. M. (2012). Structural equation modeling with Mplus: Basic concepts, applications and programming New York: Routledge. Byrne, B. M., Oakland, T., Leong, F. T L. van de Vijver, F. J. R., Hambleton, R. K., Cheung, F. M., et al. (2009). A critical analysis of cross cultural research and testing practices: Implications for improved edu cation and training. Training and Education in Professional Psychology 3, 94 105. Byrne, B. M., Shavelson, R. J., & Muthen, B. (1989). Testing for the equivalence of factor covariance and mean structures : The issue of partial measurement invariance Psych ological Bulletin 105, 456 466. Callueng, C. M., de Carvalho, M., Isobe, M., & Oakland, T. (2012). Temperament styles of children from Japan and the United States: A cross national study. Educational Measurement and Educational Review 3, 3 23. Callueng, C. M., Lee Hang, D. M., Gonzales, R. C., Ling C., & Oakland, T.D. (2011). Temperament styles of children from Samoa and the United States. Educational Measurement and Educational Review 2 18 34. Caprara, G. V., Barbaranelli, C., Bermdez, J., Ma slach, C., & Ruch, W. (2000). Multivariate methods for the comparison of factor structures. Journal of Cross Cultural Psychology, 31 (4) 437 464 Cattell, R. B., Cattell, A. K., & Cattell, H. E. P. (1993). Sixteen Personality Factor Fifth Edition Que stionnaire. Champaign, IL: Institute for Personality and Ability Testing. Chess, S. & Thomas, A. (1984). Origins and evolution of behavior disorders New York: Brunner/Mazel.

PAGE 148

148 Chess, S. & Thomas, A. (1986). Temperament in clinical practice. New York: Guilf ord Press. Cheung, G. W. & Rensvold, R. B. (2002). Evaluating goodness of fit indexes for testing measurement invariance. Structural Equation Modeling 9, 233 255. Cheung, F. M., Leung, K., Fan, R., Zang, I. X., Sun, H. F., Gan, Y. Q., Song, W. Z., et al. (2001). Indigenous Chinese personality construct: Is the five factor model complete? Journal of Cross Cultural Psychology 32, 407 433. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2 nd ed.). Hillsdale, NJ: Erlbaum. Cole, M., & Cole, S. R. (1996). The development of children (3rd ed.). New York: W. H. Freeman and Company. Costa, P. T., Jr., & McCrae, R. R. (2001). A theoretical context for adult temperament. In T. D. Wachs & G.A. Kohnstamm, (Eds.), Temperament in context (pp. 1 19 ). Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Cross, T.L., Speirs Neumeister, K.I., & Cassidy, J. C. (2007). Psychological types of academically gifted adolescents. Gifted Child Quarterly 51 (3), 285 294. CTB/McGraw Hill (1985). California Achievement Test Monterey, CA: Author. Danielsen, S. (1991). Personality type preferences of children with and without learning difficulties. Darby, L. K. (2009). Effects of shared personality types of parents and children with high functioning autism spectrum disorders on parental stress. Unpublished doctoral Davis, S.E. (2007). Effects of motivation preferred learning styles, and perception of classroom climate on achievement in ninth and tenth grade math students. Unpublished doctoral dissertation University of Florida, Gainesville. Dempsey, R.D. (1975). Cognitive style differences between gifted and average students Unpublished doctoral d issertation, United States International University, Ann Arbor, MI. Dimitrov, D. M. (2010). Testing for factorial invariance in the context of construct validation. Measurement and Evaluation in Counseling and Development 43 (2), 121 149. DiStefano, C. (2 002). The impact of categorization with confirmatory factor analysis. Structural Equation Modeling, 9, 327 346

PAGE 149

149 Eysenck, H. J. & Eysenck, M. W. (1985). Personality and individual differences: A natural science approach New York: Plenum. Fan, X. & Thomp son, B. (2001). Confidence intervals about score reliability coefficients, please: An EPM guidelines editorial. Educational and Measurement Journal 61 (4), 517 531. Flora, D. B. & Curran, P. J. (2004). An empirical evaluation of alternative methods of est i mation for confirmatory factor analysis with ordinal data. Psychological Methods 9 (4), 466 491 Fourgurean, J., Meisgeier, C., & Swank, P. (1988). The Murphy Meisgeier Type Indicator for Children: Exploring the link between psychological type preference s of children and academic achievement. Journal of Psychological Type 16, 42 46. Galen (1916). Galen on the natural forces (A. J. Brook, Trans.). Cambridge, MA: Harvard University Press. (Original work published date unknown) Galen (1992). The art of cu re Barzel, Trans.). Haifa, Israel: Maimonides Research Institute. (Original work published date unknown) Gallager, S.A. (1990). Personality patterns of the gifted. Understanding Our Gifted 3, 10 13. Goldsmith, H. H., Buss, A. H., Plomin, R., Rothbart, M. K., Thomas, A., Chess, S., et al. (1987). Roundtable: What is temperament? Four approaches. Child Development, 58, 505 529. Gorsuch, R. L. (1983). Factor analysis (2 nd ed.). Hillsdale, NJ: Erlbaum. Hambleton, R. K. (2002). The next generation of the ITC test translation and adaptation guidelines. European Journal of Psychological Assessment 17 (3), 164 172. Harrier, L. (2005). The learning styles of children diagnosed with attention deficit/hyperac tivity Disorder (ADHD) as determined by the Student Styles Questionnaire (SSQ) University, Denton. Hergenhahn, B. R. (2001). An introduction to the history of psychology (4th ed.) Belmont, CA. Brooks/Co le. Hippocrates. (1939). Hippocrates, 1 (W.H.S. Jones, Trans). Cambridge, MA: Harvard University Press. (Original work published date unknown). Hippocrates (1988). Hippocrates, 5 (P. Potter, Trans.). Cambridge, MA: Harvard University Press. (Original wo rk published date unknown).

PAGE 150

150 Hippocrates (1994). Hippocrates, 7 (W. D. Smith, Trans.). Cambridge, MA: Harvard University Press. (Original work published date unknown). Hopwood, C. J. & Donnelan, M.B. (2010). How should the internal structure of personality inventories be evaluated?. Personality and Social Psychology Review 14 (3), 332 346. Hu, L. & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1 55 Ivanova, M. Y., Achenbach, T. M., Rescorla, L. A., Dumenci, L., Almqvist, F., Bilenberg, N., et al. (2007). The generalizability of the Youth Self Report syndrome structure in 23 societies. Journal of Consulting and Clinical Psychology 75 (5), 729 738. Jennings, D.K. (2005 ). Temperament based cognitive styles of children with serious Emotional disturbance University, Denton. Joyce, D. (2010). Essentials of temperament assessment New York: Wiley. Joyce, D. & Oakland, T. (2005). Temperament differences among children with conduct disorder and oppositional defiant diso rder. The California School Psychologist 10, 125 136. Jung, C. G. (1926). Psychological types (H. G. Baynes, Trans) New York: Harcourt, Brace, & Co. (Original work published 1920) Jung, C. G. (1945). Contributions to analytical psychology (H. G. Baynes & C. F. Baynes, Trans.) New York: Kegan Paul, Trench, Trubner, & Co. LTD. (Original work published 1928) Jung, C. G. ( 1971 ). Psychological types (R.F.C. Hull, Revision of Trans. by H. G. Baynes) Princeton, NJ: Princeton University Press. (Original work published 1921) Kamata, A. & Bauer, D. J. (2008). A note on the relation between factor analytic and item response theory models Structural Equation Modeling: A Multidisciplinary Journal 15 (1), 136 153. Katona, N. & Oakland, T. (2000). The development of temperament in Hungarian children. Hungarian Journal of Psychology .1, 17 29. Kaufman, A. S., McLean, J. E., & Lincoln, A. (1996). The relationship of the Myers Briggs Type Indicator (MBTI) to IQ level and the fluid and crystallized IQ discrepancy of the Kaufman Adolescent and Adult Intelligence Test (KAIT) Assessment, 3 (3), 225 239.

PAGE 151

151 Keirsey, D. (1998). The Keirsey Temperament Sorter II Del Mar, CA: Prometheus Nemesis. Keirsey, D., & Bates, M. (1978). Please understand me: Character and temperament types Del Mar, CA: Prometheus Nemesis Book Co. Leon, C., Oakland, T., Wei, Y., & Berrios, M. (2009) V temperament styles and comparison with their United States peers. Revista Interamericana de Psicologa/Interamerican Journal of Psychology, 43, 407 415. Loehlin, J.C. ( 1998 ). Latent variable models: An introduction to factor, path, and structural analysis (3 rd ed.). Mahwah, NJ: Erlbaum. Little, T.D., Cunningham, W.A., Sharar, G., & Widaman, K.F. (2002). To parcel or not to parcel: Exploring the question, weighing the merits. Structural Equation Modeling 9 (2), 151 173. Marsh, H.W. & Hau, K.T. (1996). Assessing goodness of fit: Is parsimony always desirable? The Journal of Experimental Education 64, 364 390. McCrae, R. R. (2001). Trait psychology and culture: Exploring intercultural comparisons. Journal of Personality. 69, 819 846 McCrae, R. R., & Costa, P. T., Jr. (1997). Conceptions and correlates of openness to experience. In R. Hogan, J. Johnson, & S. Briggs (Eds.). Handbook of personality psychology (pp. 825 847). San Diego, CA: Academic Press. McGhee, R. L., Ehrler, D. J ., & Buckhalt, J. A. (2007). Five Factor Personality Inventory Children. Austin, TX: Pro Ed. Meredith, W. (1993). Measurement equivalence, factor anal ysis and factorial equivalence. Psychometrika, 58, 525 543. Miller, J. G. (1999). Cultural psychology: Im plications for basic psychological theory. Psychological Science 10 (2), 85 91. Mills, C. J. (1993). Personality, learning style, and cognitive style profile of mathematically talented students. European Journal of High Ability 4, 70 75. Mpofu, E., Oakla nd, T., & Gwirayi, P. (2010). Temperament styles of Zimbabwean early adolescents: A Rasch measurement model analysis. South African Journal of Psychology 40 (1), 44 53. Muthen, B. & Kaplan, D. (1992). A comparison of some methodologies for the factor anal ysis of non normal likert variables. British Journal of Mathematical and Statistical Psychology, 38, 171 189.

PAGE 152

152 Muthn, L. K., & Muthn, B. O. (2006). Los Angeles, CA: Muthn & Muthn. Myers, I. B., McCaulley, M. H., Quenk, N. L., & Hammer, A. L. (1998). MBTI Manual: A guide to the development and use of the Myers Briggs Type Indicator (3rd ed.). Palo Alto, CA: Consulting Psychologists Press. Myers, I. B. & McCauley, M. H. (1985). Manual: A guide for the development and use of the Myers Briggs Type Indicator. Palo Alto, CA: Consulting Psychologists Press. Myers, I. B., & Myers, P. B. (1980). Gifts differing: Understanding personality type Palo Alto, CA: Consulting Psychological Press. Nunnally, J. C. & Bernstein, I. H. (1994 ). Psychometric theory (3 rd ed.). New York: McGraw Hill. Oakland, T., Alghorani, M. A., & Lee. D. H. (2006). Temperament based learning styles of Palestinian and US children. School Psychology International 28, 110 128. Oakland, T., Banner, D., & Livin gston, R. (2000) Temperament based learning styles of visually impaired children. Journal of Visual Impairment and Blindness, January, 26 33 Oakland, T. Faulkner, M., & Bassett, K. (2005). Tempe rament styles of children from Australia and the United States Brazil Educational and Developmental Psychologist, 19, 35 51. Oakland, T., Stafford, M., Horton, C., & Glutting, J. (2001). Temperament and vocational preferences: age, gender, and racial ethnic comparisons, Journal of Career Assessment, 9, 297 314. O akland, T. & Hatzichristou, S, (2010) Temperament styles of children from Greece and the United States. School Psychology International, 31(4), 422 437 Oakland, T. Illiescu, D., Dinca, M., & Dempsey, A. (2009). Temperament styles of Romanian children Ps ihologia Sociala 22, 70 84. Oakland, T., Joyce, D., Horton, C. & Glutting J. (2000). Temperament based learning styles of male and female gifted and non gifted children. Gifted Child Quarterly 44, 183 189. Oakland, T., & Lee, D. H. (2010) Temperament s tyles of children from South Korea and the United States. School Psychology International 77 94. Oakland, T. & Lu, L. (2006). Temperament styles of child of China and the United States. School Psychology International 27, 1 92 208.

PAGE 153

153 Oakland, T., & Mata, A. (2007) Temperament styles of children from Costa Rica and the United States. Journal of Psychological Type 67, 91-102. Oakland, T., Mogaji, A., & Dempsey, J. (2006). Temperament Styles of Nigerian and U.S. children. Journal of Psychology in Africa 16, 27-34. Oakland, T., Mpofu, E., & Sulkowski, M. (2007). Temperament styles of Zimbabwe and U.S. children. Canadian Journal of School Psychology 21, 139153. Oakland, T. & Pretorius, J. (2009). Temperament styles of children from South Africa and the United States. School Psychology International 29, 627639. Oakland, T., Rizwan, M, Aftab, S., & Callueng, C. (2011). Temperament styles of children from Pakistan and the United States. School Psychology International 33 (2), 207222. Oakland, T., Singh, K, Callueng, C., & Goen, A. (2010). Temperament styles of Indian and U.S. children. School Psychology International, 32 (6), 655670. Parker, L.D. & Robinson, D.C. (1989). Type and academically gifted adolescents. Journal of Psychological Type, 17, 6689. Patterson, G. R. (1982). Coercive family process Eugene, OR: Castalia. Pinner, R. & Forlano, G. (1943). Personality tests of partially sighted children. Journal of Applied Psychology 27, 283287. Poortinga, Y. (1995). Use of tests across cultures. In T. Oakland & R. K. Hambleton (Eds.), Interpersonal perspectives on academic assessment (pp. 187-206). Norwell, MA: Kluwer. Prior, M., Sanson, A., Smart, D., & Oberklaid, F. (2000). Pathways from infancy to adolescence : Brazil temperament project 1983-2000. Melbourne, Australia: Brazil Institute of Family Studies. Quenk, N. L. (2009). Essentials of Myers-Briggs Type Indicator Assessment (2 nd ed.). New York: John Wiley & Sons. Ramrez-Esparza, N., Gosling, S. D., Benet-Martnez, V., Potter, J., & Pennebaker, J. W. (2006). Do bilinguals have two personalities? A special case of cultural frame switching. Journal of Research in Personality 40, 99120. Rothbart, M. K. (1989). Temperament and development. In G.A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 187 247). Chichester, UK: Wiley.

PAGE 154

154 Schmitt, T. A. (2011). Current methodological considerations in exploratory and confirmatory factor analysis. Journal of Psychoeducational Assessment 29 ( 4), 304 321. Schmitt, D. P., Allik, J., McCrae, R. R., & Benet Martinez, V. (2007). The geographic distribution of big five personality traits: Patterns and profile of human self description across 56 nations. Journal of Cross Cultural Psychology 38 (2), 173 212. Schuerger, J. M. (2001). 16PF Adolescent Personality Questionnaire Champaign, IL: Institute for Personality and Ability Testing. St orr, A. (1991). Jung New York: Routledge. Strafford, M. E. (1994). Ethnic differences in responding to the studen t temperament assessment record Unpublished doctoral dissertation. The University of Texas, Austin. Stafford, M. E. & Oakland, T. ( 1996). Racial ethnic comparisons of temperament constructs for three age groups using the Student Styles Questionnaire. M easurement and Evaluation in Counseling and Development, 19 (2), 100 110 Teglasi, H. (1998a). Introduction to the mini series: Impli cations of temperament for the practice of school psychology. School Psychology Review 27, 475 478. Teglasi, H. (1998b). Temperament constructs and measu res. School Psychology Review 27, 564 585. Thomas, A., & Chess, S. (1977). Temperament and development New York: Brunner/Mazel. Thomas, A., & Chess, S. (1989). Temperament and personality In G.A. Kohnstamm, J.E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 249 261). New York: Wiley. Tobacyk, J. J., Hearn, R. E., & Wells, D. H. (1990). Jungian type and California Achievement Test performance in junior high school students at high risk for dropout. Journal of Psychological Type 19, 13 20. Urbina, S. (2004). Essentials of psychological testing Hoboken, NJ:John Wiley & Sons Vandenberg, R.J. & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and r ecommendations for organizational research. Organizational Research Methods 3, 4 70. van de Vijver, F. J. R. & Leung, K. (2000). Methodological issues in psychological research Thousand Oaks, CA: Sage.

PAGE 155

155 van de Vijver, F. J. R. & Tanzer, N. K. (1997). Bias and equivalence in cross cultural assessment: An overview. European Review of Applied Psychology 47, 263 279. Wachs, & G. A. Kohnstamm (Eds.). Temperament in Context. Mahwah, NJ: Lawrence Erlbaum Associates. Wehr, G. (1971). Portrait of Jung: An illustr ated biography (W. A. Hargreaves, Trans.). New York: Herder and Herder. (Original work published 1969). Wirth, R. J. & Edwards, M. C. (2007). Item factor analysis: Current approaches and future directions. Psychological Methods 12 (1), 58 79. Yang, C., N ay, S., Hoyle, R. H. (2010). Three approaches to using lengthy ordinal scales in structural equation models: Parceling, latent scoring, and shortening scales. Applied Psychological Measurement 34 (2), 122 142. Yu, C.Y. & Muthen, B.O. (2002). Evaluation of model fit indices for latent variable models with categor ic al and continuous outcome s (Technical Report). Los Angeles: UCLA, Graduate School of Education and Information Studies. Zafran, H.A.S. (1965). A study of personality differences between blind and sighted children. British Journal of Education Psychology 35, 329 338. Zwick, R. (1987). Assessment of dimensionality of year 15 reading data. In A.E Beaton (Ed.), Implementing the new design: The NAEP 1983 1984 technical report. Princeton, NJ: Educatio nal Testing Service.

PAGE 156

156 BIOGRAPHICAL SKETCH Carmelo M. Callueng is a native of the Philippines. He obtained his Bachelor of Science in Psychology at St. Paul University Philippines (Cum Laude) in 1988 and a Master of Science in Measurement and Evaluation at De La Salle University Manila in 1993. Before his admission to the School Psychology Program at the University of Florida, Carmelo served as a faculty member and held administrative offices, including academic chair, in the Department of Psychology at De L a Salle University Manila He also served as principal investigator or co investigator in externally funded grants that included a project sponsored by the World Bank on Philippine elementary education as well as other supported research on corporate value s of Asian employees in a multi national company and cognitive styles of Filipino students. He also held an ele cted position on the Philippine Carmelo completed his internship in Psychology in the academic year 2011 2012 Psychology in Gainesville, Florida. He anticipates receiving his Master in Education and Doctor of Philosophy in School Psychology in 2012 from the University of Florida. While at the U niversity of Florida, Carmelo authored or co authored articles in quality peer reviewed journals and presented several papers at national and international conferences He holds memberships in the American Psychological Association, National Association of School Psychologists, Florida Association of School Psychologists, International Association of School Psychology, International Association of Cross Cultural Psychology, and the Psychological Association of the Philippines.

PAGE 157

157 He has received the following r esearch related awards/ scholarships : American Psychological Association D ivision 33 Research Excellence Award (2009 and 2012), the A merican Academy of School Psychology Irwin Hyman/Nadine Lambert Memorial Scholarship (2011), the Flo rida Association of School Psych ology Doctoral Graduate Studies Award (2011), the University of Florida College of Education Graduate Research Award (2012), International School Psychology Association Cal Catterall Award (2012), American Psychological Asso ciation Division 42 Graduate Research Recognition Award (2012), and the American Psychological Association Division 52 Student Poster Award (2012).