An experimental investigation of the form of information presentation, psychological type of the user, and performance w...

MISSING IMAGE

Material Information

Title:
An experimental investigation of the form of information presentation, psychological type of the user, and performance within the context of a management information system
Physical Description:
x, 125 leaves : ill. ; 28 cm.
Language:
English
Creator:
Davis, Donald L ( Donald Lamar ), 1933-
Publication Date:

Subjects

Subjects / Keywords:
Management information systems   ( lcsh )
Management thesis Ph. D   ( lcsh )
Dissertations, Academic -- Management -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1981.
Bibliography:
Bibliography: leaves 81-84
Statement of Responsibility:
by Donald L. Davis.
General Note:
Typescript.
General Note:
Vita.
General Note:
REPL*

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000294571
oclc - 07747369
notis - ABS0905
sobekcm - AA00004902_00001
System ID:
AA00004902:00001

Full Text









AN EXPERIMENTAL INVESTIGATION OF THE FORM OF
INFORMATION PRESENTATION, PSYCHOLOGICAL TYPE OF
THE USER, AND PERFORMANCE WITHIN THE CONTEXT OF A
MANAGEMENT INFORMATION SYSTEM










BY

DONALD L. DAVIS


A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA IN PARTIAL
FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1980




























Copyright 1981

by

Donald L. Davis





























TO DONNA, DONNIE, AND MARK













ACKNOWLEDGEMENTS


I wish to express my gratitude to my dissertation

advisor, Dr. Richard A. Elnicki, and to the rest of my

committee, Dr. Roger W. Elliott, Dr. H. Russell Fogler,

Dr. William M. Fox, and Dr. Marvin E. Shaw. They provided

invaluable assistance and encouragement throughout the

course of this effort. Also, I wish to express my

gratitude to my friend and colleague, William A. Shrode,

who provided so much aid and encouragement during some

rather trying times.

A special debt of gratitude is owed to my wife,

Donna, my sons, Donnie and Mark, and my mother and father,

Mr. and Mrs. Lester Tesch, for their encouragement and

sacrifice in the pursuit of our goal. It was not in vain.













TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS....................... ............. iv

LIST OF TABLES....................................... vii

ABSTRACT....................................................... ix


CHAPTER 1 RESEARCH BACKGROUND..................... 1

Introduction................... ............. 1
Literature Review............................ 4
The Minnesota Experiments................ 4
Other Studies............ ................. 10
Jungian Studies........................... 14
Summary and Evaluation of Prior Studies...... 22
Validity of the Myers-Briggs Typing
Instrument............................... 26
Purpose of the Study ......................... 28
Organization of the Dissertation.............. 29


CHAPTER 2 RESEARCH METHODOLOGY.................... 30

Introduction ................................. 30
General Model ................................ 30
Methodology.................................. 31
Variables in the Study and Research
Hypotheses ......................... .... 36
Independent Variables................... 36
Dependent Variables..................... 38
Research Hypotheses......................... 39
Experimental Design.......................... 39
Research Implications........................ 41


CHAPTER 3 EXPERIMENTAL RESULTS.................... 49

Introduction................................. 49
Results Pertaining to Hypothesis Hi.......... 49
Results Pertaining to Hypothesis H2........... 53
Results Pertaining to Hypothesis H3 .......... 56








CHAPTER 4 DISCUSSION OF RESULTS....................

Introduction.................................
Implications of Hypothesis HI.................
Implications of Hypothesis H2..................
Implications of Hypothesis H3..................
Comparison with Other Studies .................
Discussion of Post-Experimental
Questionnaire Results......................


CHAPTER 5


REFERENCES..

APPENDIX A

APPENDIX B

APPENDIX C

APPENDIX D

APPENDIX E

APPENDIX F

APPENDIX G


SUMMARY AND POSSIBLE EXTENSIONS..........


..........................................

SIMPRO LISTING..........................

INPUT ROUTINE............................

INSTRUCTIONS FOR USING SIMPRO............

EXAMPLES OF SIMPRO OUTPUT...............

TASK STRUCTURE RATING INSTRUMENT.........

POST-EXPERIMENTAL QUESTIONNAIRE.........

ASSIGNMENT TO TREATMENT GROUPS...........


BIOGRAPHICAL SKETCH...................................


Page

65

65
65
66
67
72

73


78


81

86

97

101

107

112

115

124

125













LIST OF TABLES



TABLE 1 TASK STRUCTURE RATING RESULTS ...............37

TABLE 2 FACTOR DISPLAY AND LEVELS ................... 42

TABLE 3 RESULTS BY PSYCHOLOGICAL TYPE ...............52

TABLE 4 RESULTS BY REPORT TYPE ......................55

TABLE 5 RESULTS FOR IT TYPES BY REPORT TYPE..........58

TABLE 6 RESULTS FOR ST TYPES BY REPORT TYPE......... 60

TABLE 7 RESULTS FOR IF TYPES BY REPORT TYPE..........62

TABLE 8 RESULTS FOR SF TYPES BY REPORT TYPE..........63

TABLE 9 RESULTS OF POST-EXPERIMENTAL QUESTIONNAIRE..77


vii













Abstract of Dissertation Presented to the Graduate
Council of the University of Florida in Partial
Fulfillment of the Requirements for the
Degree of Doctor of Philosophy


AN EXPERIMENTAL INVESTIGATION OF THE FORM OF
INFORMATION PRESENTATION, PSYCHOLOGICAL TYPE OF
THE USER, AND PERFORMANCE WITHIN THE CONTEXT
OF A MANAGEMENT INFORMATION SYSTEM


By


DONALD L. DAVIS

March 1981




Chairman: Richard A. Elnicki
Major Department: Management and Administrative Sciences

This study examined the effects of the psychological

type of a user and the report type provided a user of a

management information system (MIS), on user performance.

An experimental model for user performance involving

three dependent variables (cost of production, decision

time, and level of confidence) and three independent

variables (psychological type of the user, the production

environment, and report type) was used in the study. The

model was used to examine the effects on performance by

four different report types and four different psychologi-

cal types of production managers.


viii







The four report types involved the combination of

two different formats tabularr versus graphical), and two

different levels of summarization (statistically summarized

versus raw data). The psychological type of the user was

determined by combining perception and evaluation modes

obtained by the user from application of the Myers-Briggs

typing instrument. Thus, a user had a psychological type

of sensing-thinking, intuitive-thinking, sensing-feeling,

or intuitive-feeling.

The experiment was conducted using ninety-six MBA

students as subjects in a computer-simulated production

environment. Each subject was randomly assigned to a

treatment group. The subjects assumed the role of a

production manager, made decisions for levels of production

for the simulated company using reports generated by the

simulator, and received reports concerning the outcomes

of their decisions, on which they based future decisions.

The objective of the simulation was to obtain a minimum

cost for a simulated year of production.

Multivariate and univariate analysis of the experi-

mental data provided evidence of significant results due

to the experimental treatments. The results suggest

important implications for future MIS design and for

selection of production managers in existing MIS environ-

ments, given that the experimental results hold within an

actual production environment.








In the area of future MIS design, results indicated

that if an organization has no knowledge of the cognitive

type of report users, the best reports to offer in the

production environment are the tabular-raw data or

graphical-raw data reports. Where the cognitive type of

the user is known, the MIS design should offer graphical-

raw data reports to sensing-thinking users, tabular-raw

data or tabular-statistically summarized reports to sensing-

feeling users, tabular-raw data or graphical-statistically

summarized data reports to intuitive-feeling users, and

any of the four report types to intuitive-thinking users.

Given an existing MIS design, experimental results

suggest that a firm dominated by a tabular-raw data report

structure should give preference to sensing-feeling or

intuitive-feeling types for production managers. An

organization with a graphical-raw data dominated report

structure should give preference to sensing-thinking,

sensing-feeling, or intuitive-feeling types for production

managers. For a firm whose report structure is primarily

graphical-statistically summarized data, intuitive-feeling

types should be given preference as production managers.

Sensing-feeling types should be given preference for

production manager in firms whose report structure is

dominated by tabular-statistically summarized data reports.

The results also suggest that intuitive-thinking types

are incompatible with the production environment.














CHAPTER 1

RESEARCH BACKGROUND


Introduction

The proliferation of management information systems

(MIS), along with the many disappointments associated with

those systems, has generated an ever-increasing interest in

research in the area. The dissatisfaction expressed by MIS

users [6, p. 1335] may be partially attributed to the con-

fusion concerning just what an MIS is. Small and Lee state

that "the elusiveness of most MIS definitions leads natu-

rally to vague understanding and to elusive promises [33,

p. 49]. Simon leaves no doubt that a definitional problem

exists, "There is no agreement on the term management infor-

mation system" [32, p. 126]. The lack of agreement on what

is contained in an MIS has hampered research in the MIS

area [8, p. 14]. Many definitions are too general for use

in MIS research or in MIS design. As an example, Pokempner

considers an MIS to be "the most highly formalized of the

many procedures in business and government by which data

are transformed into information" [27, p. 45]. One MIS

advocate,. Bishop Walton, solved the definitional problem

by the following statement: "It is almost impossible to

define an information system, but it is easy to recognize

one" [15, p. 661.









The definition followed in this study was proposed

by Small and Lee [33]. They define an MIS as "a system

for providing information for management" [33, p. 50]

where

a system implies order, arrangement, and purpose.
Information should be distinguished from data.
Data are raw facts in isolation and do not become
information until someone has a need to know and
utilizes the data to become informed. [33, p. 50]

This definition is general enough to include all design

aspects of MIS and yet is restrictive enough to be oper-

ational.

A primary source of dissatisfaction with information

systems is that many have been designed to suit the

psychology or decision-making style of the analyst/designer,

not that of the user of the system [2, 11, 12, 29]. This

fact led Mason and Mitroff to propose the following as

one of their principles toguide MIS researchers:

Managers need "information" that is geared for
THEIR psychology and NOT to that of their design-
ers they must not only find out what infor-
mation the manager actually needs (as Ackoff [2]
points out, this is often far removed from what they
think they need), but the designers must also find
out which mode of displaying the information is most
amenable to the manager's psychology. [23, p. 485]

The psychology of the manager referred to by Mason and

Mitroff in the above quote is characterized by the Jungian

typology [26, p. 485) and implemented by the Myers-Briggs

Type indicator [26]. The decision maker processes infor-

mation by perception and evaluation. The indicator

presents two modes of perceiving and two modes of evalu-

ating what has been perceived [26, p. 1]. The two modes








for perceiving are sensation and intuition. Sensation is

perceived by use of the five senses [26, p. 1]. Intuition

is an "indirect perception by way of the unconscious,

accompanied by ideas or associations which the unconscious

tacks on to the perception from the outside." [26, p. 51].

The two modes for the evaluation process are thinking

and feeling [26, p. 52]. Thinking is a logical process.

It is impersonal and rational in nature [26, p. 52].

Feeling is a process of appreciation, subjective in nature

[26, p. 52].

Mason and Mitroff extended their argument for research

on both the presentation of information and the psychologi-

cal type of the user:

What is information for one type (person) will
definitely not be information for another. Thus,
as designers of MIS, our job is not to get (or
force) all types to conform to one, but to give
each type the kind of information he is psycho-
logically attuned to and will use most effectively.
[23, p. 478]

The view that the psychological type of the user should

be used in MIS experimentation is reinforced by Benbasat

and Schroeder:

Experiments should be designed to include both
human and MIS design variables. Most MIS theoret-
ical frameworks recognize the central importance
of the business decision maker as a determinant
of the type of information needed. Future work
should focus on instruments to describe human
characteristics and on incorporating these char-
acteristics into research designs. [4, p. 47]

Another advocate of matching reports to the user's

needs is Radford:








The degree to which output of the information sys-
tem meets a manager's needs for support in decision
making depends on two factors: (a) the nature and
form of the information presented and (b) the human
capability for absorbing, processing, and using
information. The characteristics of the human
information processing system must, therefore, be
taken into account when making decisions regarding
the form of the output from the organizational
system. Lack of attention to this requirement may
result in the output of the organizational system
being unacceptable or less valuable to the managers
concerned. [28, p. 17]

Literature Review

Much of the empirical research on this subject in the

recent past was done by members of the University of Minne-

sota's MIS department and is appropriately called "The

Minnesota Experiments" [10].

The Minnesota Experiments

The Minnesota Experiments focused on the relationship

between decision activities and information system struc-

ture. A major contribution from and an integral part of

all the experiments was the descriptive model developed

by Dickson, Chervany, and Kozar [9, p. 21] which relates

MIS variables to user performance. The model considers

user performance (P) (measured by total cost under the

user's control, time taken to make a decision, or other

measures depending on the decision environment and other

variables in the model) to be affected by the environment

of the decision maker (DE), the personal characteristics

of the decision maker (DM), and the characteristics of

the information system (CIS). The decision environment

variable (DE) is comprised of functional area (i.e., finance








production, marketing, personnel, research and development),

organizational level (i.e., strategic, tactical, opera-

tional), and the environmental measures (i.e., stability,

competitiveness, and time pressure). The decision maker

variable (DM) consists of cognitive style and other

directly and indirectly acquired attributes such as apti-

tudes and attitudes of the decision maker. The characteris-

tics of the information system variable (CIS) is composed

of the form, content, level of summarization, presentation

media, time availability, and decision aids (i.e., pre-

diction equations, EOQ models, etc.) [10, p. 918]. In

functional form, the model is expressed as:

P = f(DE, DM, CIS)

The series of experiments began with a study by

Chervany and Dickson [6] of the effects of two different

forms of information presentation on performance in a

production environment. The other experiments in the

series were based on that original study using the Dickson

et al. model and the technique of experimental gaming with

a computer-based simulator to create an artificial decision-

making environment. Experimental gaming involves the

interaction of the subjects with a computer program which

simluates the decision environment. The subjects act as

decision makers, use their decisions as inputs to the

program, and receive as outputs the simulated results of

their decisions in the form of reports [6, p. 1336].








One of the Minnesota experiments [6] was concerned

with the CIS variable alone, while four [3, 4, 20, 30]

involved the DM and CIS variables. All five impinge on

the proposed study and are discussed below.

Chervany and Dickson [6] used 22 MBA students as

subjects to study the effects of varying the level of

summarization of a report on performance. Subjects were

paired by ATGSB (Aptitude Test for Graduate Study in

Business) scores and randomly assigned to a report set.

Subjects were provided with either a statistically sum-

marized production report or a raw data (detailed) pro-

duction report set (CIS). The subjects used the reports

to reach decisions for simulated production runs. Three

practice runs and ten actual runs were made and therefore

a total of ten decisions was made by each subject. For

each run, each subject also gave a level of confidence

(1-10) for the decision and the time necessary to make the

decision. Performance was primarily measured by the total

cost (hire/fire costs + raw material costs + inventory

costs) incurred in the simulated production runs, but

decision time and confidence level were used as well.

The results of the study indicated that subjects with sum-

marized data reports performed better, but had higher

decision times and lower confidence in their decisions

than those subjects receiving raw data reports. However,

the results were not significant.









Barkin and Dickson 13] used undergraduate business

students as subjects to study the effects of cognitive

style and the effect of ordering of relevant and irrelevant

reports on using information. Cognitive style (DM) was

depicted as analytic versus heuristic, based on a 17-question

instrument titled WCFAFT 3.2. The sample consisted of 11

heuristics and 15 analytics. The subjects received summar-

ized production reports (CIS) with 198 information elements,

87 judged critical to the production decision by experts and

111 judged noncritical. There was no measure of performance.

Subjects highlighted the information items used by them in

reaching a decision. The results of the study were that

heuristics tended to select more information items than did

analytics. As well, heuristics tended to select more irrel-

evant items overall.

Benbasat and Schroeder [4] conducted an experiment to

study the effects of cognitive style (DM), format (CIS), num-

ber of reports available (CIS), and knowledge of functional

area (DM) on performance. Subjects were classified as ana-

lytic or heuristic using the WCFAFT 3.2. Reports were gener-

ated by a production simulator with ten runs as in the Dickson

and Chervany study [6], but the format, rather than the level

of summarization, was varied. Reports were either tabular

or graphical. Decision-making aids in the form of forecasts

and EOQ models were available. Exception reports were avail-

able to a treatment group on a ten-percent change in the

mean of an information item. One group had a total of eight








"necessary" reports available, while another group had

eleven additional reports available on request. Subjects

were classified as high or low knowledge on the basis of

a fourteen-item test concerning production knowledge.

Performance was measured primarily by cost (ordering +

carrying + stockout + unit production cost), but also

included decision time and number of reports requested.

Thirty-two subjects were assigned among treatment

groups on the basis of their decision-making style and

functional area knowledge level. Incentive was increased

by a monetary award. The subjects were undergraduate

students in a production course.

Two main effects significantly affected cost--format

and decision aids. Subjects receiving graphical reports

did better than those receiving tabular reports. Subjects

receiving decision aids did better than those without

decision aids. Subjects using decision aids took longer

times to make decisions. At the interactive level, subjects

who had neither graphical reports nor decision aids in-

curred the largest cost. Decision-making style as measured

by the WCFAFT 3.2 had no significant effect on performance.

As well, the number of reports available had no significant

effect on performance.

Senn and Dickson [30] conducted an experiment involving

procurement decisions (DE) with purchasing managers from

large versus small organizations as subjects (DM). A

simulated gaming environment was used, as in all the








"Minnesota Experiments". Paper versus CRT (cathode ray

tube) reports (CIS), summary versus detailed reports

(CIS) were used as treatments. This experiment was a

replication of previous experiments, but in a different

decision environment with actual rather than surrogate

subjects. Performance was measured on cost (hire/fire

costs + raw material costs + inventory costs), decision

confidence, and time.

The subjects received either a detailed output from

a line printer, a summarized output from a line printer,

or a summarized output on a CRT terminal. Results indi-

cated that no significant relation existed between

organization size and performance levels. However, CRT

users made faster decisions and requested fewer reports.

There was no significant difference in total cost perfor-

mance between groups receiving summary and detailed reports.

Another study involving DM and CIS variables was

conducted by Kozar [201. The subjects were MBA students.

The study was primarily designed to build on the Chervany

and Dickson study [6] and determine the effects of media

(CIS) and quantitative ability (DM) (as measured by the

Aptitude Test for Graduate Study in Business) in perfor-

mance measured by cost (hire/fire costs + raw material

costs + inventory costs), decision time, and confidence.

Subjects received statistically summarized reports

from a line printer or a CRT. The only significant result

was that CRT users had longer decision times than paper









report users. Quantitative ability had no effect on perfor-

mance, There was no significant difference found for cost

performance or decision confidence,


Other Studies

A study using a modified version of the Dickson et al.

model was conducted by Amador [1]. In his study, Amador

considered the effects of ordering of information items (CIS),

graphical versus tabular format (CIS), point estimates versus

interval estimates (CIS), and the number of decision entities

(CIS), where decision entities refers to "the separate pieces

of information present on a report and on which decisions are

required" [1, p. 26] on performance. Performance was measured

by cost, decision time, and choice behavior (i.e., the number

of checks performed by a subject). A computer simulation

program was used for the experiment. One hundred and sixty

undergraduate students who had completed their first undergrad-

uate statistics course were the subjects for the experiment.

The subjects were randomly assigned to each of the sixteen

treatment groups, i.e., four factors of two levels each.

The experiment consisted of a subject's using a report

to predict the occurrence of certain events. The subject

was given an opportunity to check via the simulator to

determine if the event had occurred. The event could be

"detected" only if the check was made on the simulated day

of occurrence. If the event was not detected, a cost of

five dollars was assessed. Each check cost the subject one

dollar. The objective for a subject was to minimize the









total cost of checking for the occurrence of the event

and missing the occurrence of the event itself.

Results indicated that ordering of information items

affected all three performance measures jointly. Tabular

versus graphical data also affected all three performance

measures jointly. Also, subjects with graphical reports

made more checks than those with tabular reports. Subjects

with the more convenient ordering of information items had

significantly shorter decision times. Subjects with a

"high density of decision entities" report were more sensi-

tive to report format differences than those subjects with

"low density of decision entities" reports.

The first significant field study was performed by

Lucas [21] using a descriptive model he had developed to

study the effects of several classes of MIS variables on

a user's performance. Lucas's model was unique, since it

provided for the feedback of user performance to be used

in evaluating the effect of previous user performance on

the use of the information system by the user. The var-

iables included in Lucas's model are essentially the same

as those in the Dickson et al. model.

Lucas's model considers performance to be a function

of action, analysis, personal and decision-style variables

(primarily DM in the Dickson et al. model), the quality

of the information systems (CIS in the Dickson et al.

model), and the situational factors (DE in the Dickson

et al. model). Further, Lucas argues that the performance









of the user is affected by the manner in which he uses the

information system (U). Continuing with Dickson et al.

variables for consistency in functional terms, Lucas's

model is expressed as

P = f(DE, DM*, CIS)

where DM* consists of characteristics of the decision maker

in the Dickson et al. model plus the manner in which the

user uses the information system.

In a test of the model, Lucas conducted a field study

in an organization with a computerized information system.

Using salesmen as his subjects and their total dollar book-

incs as the measure of performance, Lucas's results, in

general, conformed to those of the Minnesota experiments.

Further, his results showed that when relevant information

is provided and used, increased usage of the system in-

creases performance. On the other hand, when the informa-

tion provided is irrelevant to the decisions that must be

made, performance is negatively affected by increasing the

usage. One of the most important implications of the study,

according to Lucas, was that different personal, situational,

and decision style variables (DM*) appeared to affect the

use of the system. This is another indication that the

psychological type of the user should play a role in the

design of a system.

Lusk and Kersnick [22] report a study involving the

effects of report format (CIS), level of summarization

(CIS), user perception (DM), and report complexity (DM)





13

on performance. Report format was at two levels, tabular

and graphical, while the level of summarization was at

three levels--raw data, cumulative frequencies, and per-

centages. User perception was determined by a score on

the Embedded Figures Test (EFT), whereby subjects were

classified as high analytic or low analytic. Performance

was determined by the number of correct answers to a

twenty-question instrument involving simple arithmetic.

The answers were derived by the subject using information

from one of the experimental reports. An example of the

questions is as follows: "How many of all professional

sampled were accountants who earned $20,000 or less?"

[22, p. 790] Subjects were 219 undergraduate students.

No information was given on the class level of the students.

The EFT was administered to the subjects as they met for

their regular classes. Each subject was given all of the

following types of reports to rank for complexity:

tabular-raw data, tabular-percentages, graphical (frequency

histogram)-raw data, graphical (cumulative frequency)-

raw data, graphical (cumulative frequency)-percentage.

In a second session one week after the first, subjects

were given a particular report type and the twenty-question

test. Although the authors did not comment on the possible

effect, the test was time constrained. Monetary prizes

of $10 and $5 were given for best performances.

The results showed no significant effect due to psycho-

logical type. Tabular reports were perceived as less

complex by the subjects and resulted in the best performances.









Jungian Studies

Some studies have used the Jungian psychology of types

[18], as implemented by the Myers-Briggs typing instrument,

to study the effect of cognitive type. The indicator

following the Jungian typology presents two modes of per-

ceiving and two modes of evaluating what has been perceived

[26, p. 1].

As described earlier in this chapter, the two modes

for perceiving are sensation and intuition [26, p. 51].

Sensation is perceiving primarily by use of the five senses

[26, p. 51]. Intuition, on the other hand, is an "indirect
perception by way of the unconscious, accompanied by ideas

or associations which the unconscious tacks on to percep-

tion from the outside" [26, p. 51]. The appendages by the

unconscious can range from the simplest hunch to the scien-

tific discovery. The individual who is a sensation type

prefers detail and objective hard facts, while the intuitive

type prefers possibilities and totality or Gestalt in the

perception process [26, p. 56].

The two modes for the evaluation process are thinking

and feeling [26, p. 52]. Thinking is a logical process.

It is impersonal and rational in nature [26, p. 52].

Feeling, in contrast, is a process of appreciation, sub-

jective in nature [26, p. 52]. The thinking type is char-

acterized by abstract true/false judgement, while the

feeling type is characterized by evaluations in the good/

bad, pleasant/unpleasant vein [26, p. 52]-. The modes








tend to be mutually exclusive, and the two modes for

perception are independent of the two modes of evaluation

[26, p. 53]. Thus, the four modes can be partitioned

into psycholocial types [26, p. 53]. The four levels

of the psychological-type variables are [26, p. 53]:
1. Sensation-Thinking (ST)
2. Sensation-Feeling (SF)
3. Intuition-Thinking (IT)
4. Intuition-Feeling (IF)

This instrument has been used extensively in many fields.

Furthermore, it has come to be used by researchers investi-

gating managerial behavior. McKenney and Keen [24] used

the Myers-Briggs instrument to type managerial decision-

making styles, as did Hellriegel and Slocum [17], Smith

and Urban [34], and Mitroff and Kilmann [25].

McKenney and Keen's study [24], although not primarily

concerned with MIS's, examined the cognitive process in-

volved with the information processing and decision making

by managers.

They developed a model of cognitive style with two

dimensions, information gathering and information evaluation

[24, p. 80-81]. Information gathering is a perceptual

process involving the inputing, filtering, and categorizing

of information [24, p. 80]. Information evaluation refers-

to the process of problem solving [24, p. 81]. Information

gathering is further divided into two separate and indepen-

dent classifications, perceptive and receptive [24, p. 80].

"Perceptive individuals bring to bear concepts to filter

data; they focus on relationships between items and look








for deviations from or conformities with their expectations"

[24, p. 80]. Receptive individuals are "more sensitive to

the stimulus itself. They focus on detail rather than on

relationships and try to derive the attributes of the

information from direct examination of it instead of

fitting it to their precepts" [24, p. 80].

Information evaluation is also divided into two

classifications, systematic and intuitive. Systematics

"tend to approach a problem by structuring it in terms of

some method which, if followed through, leads to a likely

solution." [24, p. 81] Intuitives "usually avoid committing

themselves in this way. Their strategy is more one of

solution-testing and trial-and-error. They are much more

willing to jump from one method to another, to discard

information, and to be sensitive to cues that they may

not be able to identify verbally." [24, p. 81].

McKenney and Keen reported the results of three

studies conducted using their model [24]. In the first

study, twenty MBA students, previously tested and showing

strong cognitive styles, participated in a study that made

use of a "cafeteria" set of 16 problems [24, p. 84]. The

subjects chose any five problems to answer and were invited,

but not required, to describe their actions while involved

in the solution. Results indicated that systematics de-

veloped algorithms or methods to solve the problems, while

intuitives attacked the problems and tried something to

see where it led them [24, p. 84]. The method of the









intuitive "generally showed a pattern of rapid solution,

abandoning lines of exploration that did not seem prof-

itable" [24, p. 84]. Further, "systematics preferred

program-type problems, while intuitives like open-ended

ones" [24, p. 84J,

In another study with the same subjects, McKenney

and Keen compared the results of their first study with

Myers-Briggs scales [24, p. 84]. A significant relation-

ship was found between the McKenney and Keen model and

that of the Myers-Briggs. The strongest relationships

were found between Jungian thinking types and McKenney

and Keen systematics, and between Jungian feeling types

and McKenney and Keen intuitives [24, p. 84].

The third study by McKenney and Keen used eighty-four

MBA students as subjects in examining the relationship

between cognitive style and career choice. The career

preferences between systematic and intuitive subjects were

compared [24, p. 85]. Results indicated that systematic

students preferred administrative, military, production,

planning, control, and supervisory careers, while intuitives

were oriented toward careers in psychology, advertising,

library science, teaching, and the arts [24, p. 85].

Mitroff and Kilmann use the four Jungian psychological

types to describe four forms of management science [25].

The sensation-thinking (ST) form emphasizes precision,

control, specific, impersonal analysis and logic, i.e.,









quantitative analysis [25, p, 19], The intuitive-thinking

form of management science (IT), on the other hand, stresses

conceptual analysis, a qualitative type of analysis [25,

p. 19]. The two other Jungian types of management science

are qualitative as well but are slightly different [25, p. 20].

Both the intuitive-feeling (IF) and the sensation-

feeling (SF) types rely on subjective and value criteria

for analysis rather than on impersonal, logical roles [25,

p. 20]. "Rather than attempting to find the common theme

or character of some set of phenomena, the feeling function

strives to generate differences [25, p. 20]. The major

difference between the IF and SF types of management

science approach is that SF types strive for some precision

while IF's can be abstract and loosely defined [25, p. 20].

To support their classifications, Mitroff and Kilmann

describe a study which explores "whether individuals with

different psychological types do in fact have different

views of organizational design and what constitutes an

ideal organization" [25, p. 20].

Three different groups of managers were classified

according to their psychological types. The subjects were

then asked to write a short story expressing their concept

of an ideal organization. Upon completion of the story,

each subject was placed in a discussion group based on

psychological type. Each group was told to organize them-

selves as they saw fit, consider the story of each subject








in the group, and then compose a group story which best

expressed the groups concept of an ideal organization [25,

p. 21].

Each group of managers was tested separately. One

group consisted of 25 middle-to-high level managers of

business organizations in the Pittsburgh, Pennsylvania,

area [25, p. 22]. The other two groups were middle level

supervisors in the Pennsylvania State Department of Public

Assistance [25, p. 22].

Results indicate that

(1) There is a remarkable and very strong similarity
between the stories of those individuals who
have the same personality type (e.g., ST)

(2) There is a remarkable and very strong difference
between stories of the four personality types
[24, p. 21]

Hellriegel and Slocum [17] developed a composite

model of managerial decision-making styles based on the

Jungian psychology of types. According to this mode,

SF managers

are interested in facts that can be collected
and verified directly by the senses. They
approach these facts with personal and human
concern because they are more interested in facts
about people than about things. When asked to
write a paragraph or two on their perception of
an ideal organization, these individuals often
describe an organization with a well defined
hierarchy and set of rules that exist for the
benefit of members and society. The ideal
organization would also satisfy member needs and
enable them to communicate openly with one another.
[17, p. 35]








While an IF manager would

rely primarily on intution for purposes of
perception and feeling for purposes of decision
making. These managers focus on new projects,
new approaches, new truths, possible events and
the like. They approach these possibilities in
terms of meeting or serving the personal and social
needs of people in general. Intuitive feeling
types avoid specifics and focus instead on broad
themes that revolve around the human purposes of
organizations, such as serving mankind or the
organization's clientele. The ideal organization
for these individuals would be decentralized, with
flexible and loosely-defined lines of authority
and few required rules and standard operating
procedures. [17, p. 35]

The ST manager

emphasizes external factual details and specifics
of a problem. The facts of a problem are often
analyzed through a logical step-by-step process
of reasoning from cause to effect. This manager's
problem-solving style tends to be practical and
matter-of-fact. When asked to describe his ideal
organization, this individual often describes an
extreme form of bureaucracy, characterized by its
extensive use of rules and regulations, a well-
defined heirarchy, emphasis on high control,
specificity and certainty, and its concern with
realistic, limited and short term goals.
[17, p. 35]

The fourth composite, IT managers

tend to focus on possibilities, but approach them
through inpersonal analysis. Rather than dealing
with the human element, they consider possibilities
which are more often theoretical or technical.
These managers are likely to enjoy positions which
are loosely defined and require abstract skills,
such as long-range planning, marketing research
and searching for new goals. The ideal organi-
zation for these individuals would be impersonal
and conceptual. Goals of the organization should
be consistent with environmental needs (such as
pure air, clear water and equal opportunity) and
the needs of organizational members. However,
these issues are considered in an abstract and
impersonal frame of reference. [17, p. 35]








One of the most recent studies is that accomplished

by Smith and Urban [34]. The study examined the relation-

ship between personality, orientation, measured by the

Myers-Briggs Type Indicator, and differences in information

processing in an ambiguous and an unambiguous information

sequence.

In the study the subjects were grouped by personality

type according to Myers-Briggs typing. The groups consisted

of 25 sensing-thinking types (ST), 22 sensing-feeling types

(SF), 18 intuitive-feeling types (IF), and 15 intuitive-

thinking types (IT) randomly assigned to two difference

classes. Comparisons were made between the groups in terms

of information processing in an ambiguous and unambiguous

information sequence. The experiment involved the imaginary

drawing of poker chips from bags containing different colors

of chips. The subjects were given a sequence of 10 draws,

an ambiguous sequence where there was no definite pattern

or predominate color and an unambiguous sequence where

each successive draw established color pattern. The

investigators found that differences in performance on

the task was quite significant between IT and IF types.

Comparisons between the 10 highest thinking and 10 highest

feeling types gave significant differences. The same

comparisons for the most pronounced sensing and intuition

types were also significant.








Summary and Evaluation of Prior Studies

From the foregoing review of pertinent literature,

one can see that empirical research in the MIS area is

relatively new. The lack of MIS research led to Mason

and Mitroff's development of a framework for MIS research

and a call for empirical research in the area, to include

the psychological type of the decision maker and report

construction as design variables [231.

The Minnesota experiments were among the first to use

a descriptive model which included the performance of the

user of an MIS as the dependent variable while using

characteristics of the user and report characteristics as

independent variables [10]. The Minnesota experiments

were empirical studies relying on a computer program to

simulate the pertinent decision environment, generally

employing surrogates in the role of decision-makers [3, 4,

5, 9, 10].

The first field study involving actual users of an

MIS rather than surrogates as subjects was performed by

Lucas. Lucas's descriptive model, basically the same as

that used in the Minnesota experiments, added utilization

of the system as a factor in user performance and involved

salesman as subjects [21].

Results from previous studies concerning the effects

of format and level of summarization have been somewhat

contradictory. Format has had a significant effect in all

the studies mentioned [1, 4, 22]. However, reported results








for effects from level of summarization have been mixed.

Amador [1] found significant effects on performance due to

the level of summarization, but Chervany and Dickson [6],

Senn and Dickson [30], and Lusk and Kersnick [22] did not.

Amador's results were found when he considered effects on

the performance (dependent) variables jointly, instead of

univariate effects on a single independent variable. This

may be the key, as Amador argues in his study [1, p. 41].

Three [3, 4, 22] of the previously mentioned studies

examined the effect of cognitive type of the decision maker

on performance. However, only in the Barkin and Dickson

study [3] was there any significant effect reported, and

then only in the utilization of the system, not on perfor-

mance. This seems almost unbelievable to the author,

since it appears intuitively obvious that there are

different "best" ways of reporting information for managers

of different cognitive types, or at least better ways than

providing only one type of report for all managers.

Further, appeal to experts indicates that there should be

better methods of designing information based on the

psychological types of managers.

Cognitive type, cognitive style, or psychological type

are all terms used interchangeably to refer to "an indi-

vidual's way of performing 'perceptual and intellectual'

activities" [35, p. 90]. Note the two processes contained

in the definition. Perceiving and intellectualizing are

recognized as two distinct dimensions. Further, in their










information system processing model, McKenney and Keen

define cognitive style as "consistent modes of thought"

which

can be classified along two dimensions, infor-
mation gathering and information evaluation .
Information gathering relates to the essentially
perceptual process by which the mind organizes the
diffuse verbal and visual stimuli it encounters.
The resultant "information" is the outcome of a
complex coding that is heavily dependent on mental
set, memory capacity, and strategies--often uncon-
scious ones--that serve to ease "cognitive strain".
Of necessity, information gathering involves reject-
ing some of the data encountered and summarizing
and'categorizing the rest .. Information
evaluation refers to processes commonly classified
under problem solving. Individuals differ not only
in their methods of gathering data, but also in their
sequence of analysis of that data. [24, pp. 80-81]

There are at least two probable reasons for the lack

of significant effect due to cognitive type in the reported

studies--use of an ineffective instrument for measuring

cognitive type, and lack of consideration of the joint effect

on correlated dependent variables.

Two of the studies [3, 4] used an instrument, WCFAFT,

developed by the MIS Department at the University of Minne-

sota, to determine psychological type. The instrument has

been characterized as measuring only "planned versus spon-

taneous" preferences in a validity study by Zmud [33]. In

that study, WCFAFT and MBTI were administered to forty-eight

MBA students. WCFAFT results were compared with MBTI results.

WCFAFT correlations with MBTI dimensions were 0.25 for Sensing-

Intuition, 0.12 for Thinking-Feeling, 0.06 for Extroversion-

Introversion, and 0.78 for Judging-Perceptive [38, p. 1089].









Zmud's results led him to believe that "any MIS cognitive

style implications from the Minnesota experiments using

WCFAFT should probably not be extended beyond the spon-

taneous versus planned dichotomy [38, p. 1090]. It is

apparent that the types of simulated conditions involved

in the Minnesota experiments (Production and Purchasing

Simulation) did not lend themselves to discriminating

strategy differentiation, and therefore cognitive type, as

measured by the WCFAFT, would not be expected to have a

significant effect on performance.

Another of the studies used the Embedded Figures Test

[22]. This test is a perceptual instrument developed by

Witkin et al. [36]. Lusk and Kersnick appeal to an untested

suggestion by Witkin et al. [36] that the EFT's dimensions

can extend beyond the perceptual process. They extended the

measurement to the process of evaluation as well [22, p. 792].

This is stretching the power of suggestion. The appropriate

conclusion should have been that the process of perception

alone showed no significant effect on performance. In

addition, the EFT has been shown to have a sexual bias

[37, p. 297]. Lusk and Kersnick did not report consideration

of this additional source of variance.

It appears that the two Minnesota studies [3, 4] using

cognitive type as an independent variable suffered from the

use of a weak instrument to measure the construct, while the

Lusk and Kersnick study examined only perception with a some-

what biased instrument. As mentioned previously, Mason









and Mitroff suggest using the Jungian typology operation-

alized by the Myers-Briggs instrument [26, p. 472].


Validity of the Myers-Briggs Typing Instrument

In the present study, the primary concern with validity

of the Myers-Briggs instrument is with the measures of

perception and evaluation, i.e., cognitive types. Researchers

[19, 23, 25, 34] have been able to use the perception and

evaluation of the MBTI because of the results of intercor-

relation studies which show perception (SI) and evaluation.

(TF) to be independent of one another [5, p. 462]. As

measured by phi coefficients, the intercorrelations of SI

and TF range from -.02 to .07 in three separate studies

[5, p. 463].

"The validity of the instrument is dependent on how well

it measures what it was intended to measure" [5, p. 467].

Three types of validity have been examined: content validity,

construct validity, and predictive validity [5, p. 467].

Content (face) validity is concerned with how adequately

the domain of the characteristic is captured by the measure

[7, p. 256]. Several investigators have conducted studies

on content validity of the Myers-Briggs. These studies have

been detailed by Carlyn [5]. The author reported studies in

which the item content of the MBTI had been examined. Carlyn

concluded that the Sensing-Intuition and Thinking-Feeling

scales "seem largely consistent with their corresponding

conceptual definitions" [5, p.469].









Predictive validity is concerned with the usefulness

of the measuring instrument as a predictor of some other

characteristic or behavior 17, p. 256]. Four studies are

listed by Carlyn on the predictive validity of the MBTI [5,

p. 468]. "The studies suggest that the instrument

has moderate predictive validity in certain areas" [5, p.

469].

Construct validity is concerned with what psycholog-

ical property or properties can explain the variance of the

instrument. It is concerned with the relationship between

variables [7, p. 258]. As reported by Carlyn, several

researchers used factor analysis to investigate the relation-

ship between the constructs measured by MBTI and constructs

measured by other instruments such as the Allport-Vernon-

Lindzey Study of Values [5, p. 469]. Carlyn concludes that

"the numerous studies of construct validity suggest

that the individual scales of the Myers-Briggs Type Indica-

tor measure important dimensions of personality which seem

to be quite similar to those postulated by Jung" [5, p. 469].

Reporting on the reliability of the MBTI as measured by

internal consistency and stability, Carlyn states that in-

ternal consistency studies have usually produced acceptable

reliabilities, 0.75 to 0.87 to SN and 0.69 to 0.86 for TF,

using Cronbach's coefficient Alpha [5, p. 465]. However,

studies of stability show some inconsistencies [5, p. 467].

Carlyn suggested that the inconsistencies were functions of

age or occupation. Carlyn concluded that "the Indicator







28

appears to be a reasonably valid instrument which is poten-

tially useful for a variety of purposes" [5, p. 472].


Purpose of the Study

Motivated by the relative scarcity of empirical research

dealing with Management Information Systems, by the lack of

significant findings, and by conflicting evidence from prev-

ious studies on the effects of the cognitive type of the

user and report type on user performance, the purpose of

the present study is to investigate and evaluate the effects

of Jungian Psychological types (ST, SF, IT, and IF) and

information presentation (format and level of summarization)

on user performance within the context of management infor-

mation systems. The cost-benefit relationship involved in

adapting reports to various psychological types is not con-

sidered in the study. The need for the research was detailed

earlier [4,23, 28], along with pertinent research on this

area.


Organization of the Dissertation

Chapter 2 contains the research methodology and impli-

cations of the results of the study. First, the general

model used to guide the study is presented, along with a

discussion of the specific methodology used for the study.

Next, the three research hypotheses are presented and

discussed. Finally, the experimental design for the study

is presented, followed by a discussion of research implica-

tions.










In Chapter 3, the experimental results of the study

are detailed. The implications of the results are dis-

cussed in Chapter 4. Also in Chapter 4, the experimental

results of this study are compared with the results of

previous studies, and the results of the post-experimental

questionnaire are discussed.

Chapter 5 is the final chapter of the dissertation.

It contains a summary of the study and possible extensions

for future research in the area of MIS design.













CHAPTER 2

RESEARCH METHODOLOGY


Introduction

As mentioned in Chapter 1, this study examines the

relationship between the psychological type of the user of

a report (implemented by the Myers-Briggs instrument), the

format of the report, the level of summarization of the

report, and their effects on performance (measured by cost,

decision time, and level of confidence in the decision).

The experiment was guided by a general model based on one

developed by Amador [1] from the tiickson et al. and Lucas

models.


General Model

Amador's model expresses user performance (P) as a

function of report format (F) and level of summarization

(L), given a particular decision environment (DE), decision

maker (DM), and other characteristics of the information

system (CIS'). Or, in functional form:

P = f(F, L I DE, DM, CIS')

Therefore, the model used to guide the present study is:

P = f(F, L, T I DE', DM', CIS')

Or, user performance (P) is a function of the report format

(F), level of summarization (L), and psychological type of








the user (T), given a production decision environment (DE'),

other characteristics of the decision maker (DM'), and

other characteristics of the information system (CIS').

Other characteristics of the decision maker (DM') are

randomly distributed over the sample population while other

characteristics of the decision environment (DE'), and the

other characteristics of the information system (CIS') are

held constant by the experimental structure and procedures.


Methodology

The experimental gaming methodology used in the

Minnesota experiments, as described earlier, was used in

the present study. Each subject, classified by psychologi-

cal type, acted as a production manager. The production

simulator is:

a computer program which simulates the production
operations of a manufacturing organization. The
program allows one or more users to interact with
the simulator, make decisions regarding production,
and receive the results of those decisions as though
a production cycle had actually taken place. The
number of cycles simulated is determined by the
user. [9, p. 916]

The specific decision environment (DE') is one where

the subjects, acting as production managers, make a number

of sequential decisions based on reports generated by the

productions simulator (CIS'). The production simulator

used in this study is entitled SIMPRO.








SIMPRO (Appendix A) is an interactive computer program

written in an extended FORTRAN to simulate the production

function in a single-product organization. SIMPRO is

adapted from UNISIM, a simulation program developed by

Roy D. Harris and Michael J. Maggard [16, pp. 197-220].

The purpose of SIMPRO is to allow an individual to

assume the role of production manager in a simulated pro-

duction decision environment. The key tasks faced by the

individual assuming the role are as follows.

1. Analysis of the demand faced by the firm.

2. Familiarization with the inventory and production
costs experienced by the firm.

3. Development of a production plan for the five-
week production period with the objective of
minimizing the firm's total costs.

4. The input of production decisions for the weeks
production run.

5. Evaluation of the operating results. (To balance
production fluctuation, finished goods inventory
and stock-out costs.)

The firm produces a single product from raw materials

obtained from raw materials inventory. The finished pro-

duct is either shipped to a customer or placed in finished

goods inventory to be shipped at a later date, as shown

below.





















The following assumptions affect the production system:

1. The firm has an unlimited supply of raw materials.

2. Actual demand for the product averages 100,000
units per week; however, demand is uncertain
and fluctuates. The range is from 79,000 to
181,000 units per week. Further, demand has a
seasonal pattern.

3. The capacity of the production system is 180,000
units per week.

4. Finsihed goods inventory is 240,000 units. For
any finished goods inventory in excess of
240,000 units, there is an extra cost.

5. When production levels are changed from one
week to the next, there are hire-fire costs,
labeled fluctuation costs.

6. When actual demand exceeds the current week's
production plus finished goods inventory a
shortage or stock-out results. The shortage
is carried over into the next and succeeding
weeks until satisfied.

The total cost which a manager attempts to minimize

is composed of the following:

1. Production fluctuation cost is a cost incurred
by changing the level of production from the
previous week. The cost of $0.10 per unit
change in production.








2. Regular inventory holding cost is a unit cost
as well. For each unit held in finished goods
inventory up to and including the 240,000th
unit, there is a $0.04 cost. Inventory at the
beginning of the first week is zero.

3. Extra inventory holding cost is incurred at the
rate of $0.08 per unit for any number of
finished goods inventory greater than 240,000
units.

4. Stock-out costs of $0.25 per unfilled unit are
incurred when current demand plus previous
unsatisfied demand cannot be met with current
production and finished goods inventory. The
unfilled units are back-ordered and the cost
is incurred in succeeding week until backorders
are satisfied.

5. Total cost is the sum of the above four costs,
i.e.,

Total Cost = Production fluctuation cost

+ regular inventory holding cost

+ extra inventory holding cost

+ stock-out cost.

The object of the game is to minimize the total

cost. This objective can be met by balancing the four

costs--that is, by incurring some of each cost to avoid

excessive cost occurring in any one area.

The input consists of the manager's social security

number, amount of time (in minutes) which the manager took

to make the decisions, the level of confidence the manager

has in the production decisions (from 1-10), and the

number of units to be produced for five weeks. All input

is from an interactive terminal, and the input is edited

for validity. The input routine is included as Appendix B,

and instructions to the user are given in Appendix C.








The program produced four output versions which vary

in format and level of summarization but not in content.

Examples of the four outputs are given in Appendix D.

To determine the extent to which the decision task

was too simple or too complex to affect cognitive differences

in decision making, the task was rated by ten production

experts to determine task structure as operationally de-

fined by Shaw [31] and used by Fiedler [13]. The task

structure dimensions are:

1. Goal clarity: the degree to which the require-
ments of a task are clearly stated or known
to the people performing the task.

2. Goal-path multiplicity: the degree to which the
problems encountered in the task can be solved
by a variety of procedures.

3. Decision verifiability:.the degree to which the
correctness of the solutions or decision typically
encountered in a job can generally be demonstrated
by appeal to authority, by logical procedures, or
by logical feedback.

4. Solution multiplicity: the degree to which there
is generally more than one "correct" solution
involved in the task. [31, p. 10]

Each expert rated each dimension on a five-point

Likert-type scale (see Appendix E). Interrater agreement

on all four scales was determined by use of Kendall's

coefficient of concordance [14, p. 263]. The task was

also rated by the subjects. As shown in Table 1, the

experts were in close agreement on rating the task, w = .86.

Also, the experts' mean rating for goal priority was 4.9

with standard deviation of 0.03, indicating that the

experts felt the production task goals were clearly stated








to a high degree with little variability. Further, with

means of 3.5 for goal path multiplicity with standard

deviation of 0.16, 3.7 for decision verifiability with

0.41 standard deviation, and 3.2 for solution multiplicity

with 0.38 standard deviation the experts, on the average,

indicated from some to a great degree that the problems

experienced in completing the production task can be solved

by using a variety of procedures, that the correctness of

the decisions can be verified in an organizational setting,

and that there is more than one correct solution involved

in the task, all with relatively little variability.

Therefore, the task was considered adequate to represent

the production environment.

Table 1 shows that the subjects were also in relatively

close agreement in rating the task, w = .70. Also, the

subjects, on the average, gave each dimension a rating of

between some and a high degree with little variability,

which is not inconsistent with the ratings provided by the

experts.

Variables in the Study and Research Hypothesis

The experiment was designed to explore the relation-

ship between three independent variables and three dependent

variables.

Independent Variables

The three independent variables are:

1. A format variable (F) which was analyzed at
two levels, tabular form and graphical form.













TABLE 1

TASK STRUCTURE RATING RESULTS


MEAN SCORE


STANDARD
DEVIATION


Goal Clarity


Experts Subjects
4.9 3.8


Experts Subjects
0.03 0.02


Goal-Path Multiplicity 3.5

Decision Verifiability 3.7

Solution Multiplicity 3.2



Expert w = .86, n = 10


Subject w = .70, n = 96


3.2

3.2

3.1


0.16

0.41

0.38


0.16

0.21

0.16








2. A level of summarization variable (L) which was
analyzed at two levels also, in conjunction with
the format variables. The two levels for the
level of summarization variable are raw data and
statistically summarized data (mean and standard
deviation).

3. Psychological type (T) variables, as specified by
Mason and Mitroff [23], were used. The psycholog-
ical type classification is based on Jung's theory
of types [10]. The basic premise of the theory
is that much apparently random variation in human
behavior is actually orderly and consistent. The
variation is due to certain basic differences in
the way people prefer to use perception and
judgement. Perception (G) includes the processes
of becoming aware objects, people, occurrences,
or ideas. Judgement (E) includes the processes
of reaching conclusions about what ;has been
perceived. If managers differ systematically
in what they perceive and in the conclusions they
reach, then it should be possible to type them
by the way they perceive and the way they evaluate
the results of their perceptions.

The Myers-Briggs Type Indicator [26, pp. 12-50] was

used to determine the psychological type of the subject.

Dependent Variables

1. Cost performance is a relatively unambiguous
measure of performance. It is also one of the
most commonly used [1, 4, 6, 9, 30]. For this
study, the total cost of production was used to
measure performance, where total cost of produc-
tion is the sum of production, fluctuating costs,
regular and extra inventory costs, and stock-out
costs.

2. Decision time performance was measured by the
time elapsed from the time a subject received
the reports to the time the subject submitted
the decisions for the following week.

3. Decision confidence was measured by a ten-point
scale, ranging from one to ten. "1" indicated
no confidence, while "10" indicated utmost
confidence in the decisions made for that week.
Subjects were asked to express their confidence
at the time their decisions on production
quantity were made.








Research Hypotheses

The proposed research was based on three hypotheses:

HI: The psychological type of the user will have
an effect on the decision activity of the user.
The performance of the users will vary by
psychological type as follows:

Highest 1. IT Intuitive-Thinking type
2. ST Sensation-Thinking type
3. IF Intuitive-Feeling type
Lowest 4. SF Sensation-Feeling type

H2: User performance will be affected by the report
format and by the level of summarization of the
data, as follows:

Highest 1. Graphical, statistically sum-
marized
2. Graphical, raw data
3. Tabular, statistically summarized
Lowest 4. Tabular, raw data

H3: Users of given psychological types will perform
better with reports of one combination of format
type and level of summarization than with any
others, as follows:

Psychological Type Report

IT Graphical, SSD
ST Tabular, Raw Data
IF Graphical, SSD
SF Tabular, Raw Data


Experimental Design

A 24 factorial design with perception (G) at two

levels, evaluation (E) at two levels, and format type (F)

and level of summarization (L) at two levels was used for

collecting and analyzing the data, as illustrated in Table

2. A total of 96 MBA students were selected by psycholo-

gical type, using the Myers-Briggs Type Indicator. Each

block consisted of four treatment groups (cells). Each








treatment group consisted of six subjects of the appropriate

psychological type, randomly assigned (see Appendix G).

The random assignment permits the assumption that all

other decision-maker characteristics (DM') are randomly

distributed among treatment groups. Each group within a

psychological type received a treatment consisting of format

type and level of summarization of data, as illustrated in

Appendix D.

The experiment was conducted over a period of two

weeks. Each subject made 13 decisions (13 runs). The first

three runs were used to familiarize the subjects with the

game, negate the learning effect, and stabilize the gaming

process. The subjects completed the experiment in one

session at the computer terminals. Each subject was

allowed all the time needed to complete the experiment. At

a time scheduled by the subject, he reported to the com-

puter laboratory and received the instructions for his

role as operations manager (Appendix C). The subject was

allowed to study the instructions and to ask questions.

When the subject was satisfied that he understood the

process, he began the experiment. The subject ran the

game in the computer laboratory using a CRT (cathode ray

tube) computer terminal. A laboratory assistant was present

to help the subject start the simulation, overcome any

technical problems with the computer, get the printed report

from the line printer, and assure that no confounding

variables, such as interruptions, or assistance from other










individuals, were introduced into the experimental process.

(See Appendix C for details of running the game.) A $50

prize was awarded for the lowest cost, $25 for the next

lowest cost, and $15 for the third lowest cost. Also,

T-shirts with the slogan "Jungian Types Have PERSONALITY"

were awarded to each participant.

Multivariate analysis of variance (MANOVA) was conducted

on the results obtained for the three dependent variables.

ANOVA was used for univariate analysis and the Duncan's

Multiple Range Test (DMRT) was used to determine significant

differences between levels of independent variables.

Following the experiment, a post-experimental question-

naire was administered. The subject was asked (1) what kind

of report he would prefer, (2) what kind of additional in-

formation, if any, he would have preferred, and (3) his

evaluation of the experimental task (see Appendix F).


Research Implications

The basic questions suggested by the three research

hypotheses and their implications are discussed below. The

first research hypothesis (Hl) addresses the question of

whether the way in which the user perceives and evaluates

what was perceived, as implemented by the Myers-Briggs Type

Indicator, generally affects the decisions of the user. The

hypothesis was examined by comparing the performance measure

of the subjects by psychological type. The anticipated















ONNW
CM 4J 0
34 24-0


4-i M4 c
i-H WN 0

3 .----1 ou

H CM C0



C14
O N W
crl0 M (






N >3 ca ca 4 *1

T () r i a Wc
FZ 0 M 0 0
HP C 4 0 o C i




4 .0 CN w

a) (a z C 0 wC
a) a) r-i =u C1 co 3 -H
0 0 r-3 Z CU< U 4









IW E-4 r-4 Cl C -
W 0 o 0
) i M c ) r4
o 0 CU O -4
4 C r W44

S) 0 M >

E-4 CD 0 E- 4


0 4O \N cc 0

4 ca M o 4 a a 0
0 (U B0 44 N 00


Cd C P
- 1- --4--- 3 f
44 C0 r4 4 O (B 4

HU H W H d C U
0* f H *W 0 0 0) u C-
z 14 0 CNN'- 54 rC
1-4 -H 0 0 w >
cl CO *rl U C UU








Sze 0- o4 N
4-W
HC r. l a





SP E-4 0














S H i c a E o 0o
u O ZC C U H
Z4 0U 4 r) v N
u0 44 0 4 w 0 (B 0
co r. 44 (3 1g zI1 >








en E-< oda,
------------ U FK il 0t M









performance ranking from highest performance to lowest per-

formance was as follows:

Highest 1. IT Intuitive-Thinking type
2. ST Sensation-Thinking type
3. IF Intuitive-Feeling type
Lowest 4. SF Sensation-Feeling type

The explanation for these expectations is found in the

description of the four psychological types [26, pp. 54-55]

and the nature of the decision tasks they face.

The intuitive-plus-thinking type (IT) uses intuition

for perception but teams it with thinking. This type focuses

on possibilities, but approaches them with impersonal analy-

sis. "Often the possibility they choose is a theoretical,

technical, or executive one, with the human element subor-

dinated" [25, p. 55]. Given the operational nature of the

production game, an individual who perceives in a Gestalt

fashion can grasp the key issues of the problem, operate on

them in an impersonal analytical manner, and thus should

perform well.

The sensation-plus-thinking type (ST) relies primarily

on sensing for purposes of perception [26, p. 54]. But he

uses thinking for purposes of evaluation [26, p. 54]. The

ST type focuses on facts which can be collected and verified

by the senses, and makes decisions based on those facts by

impersonal analysis [26, p. 54]. Therefore, in the opera-

tional environment, individuals of this type should perform

well, but not quite as well as those who perceive the "whole"

and evaluate it in a logical manner, as the IT types do.








Those in the intuitive-plus-feeling category (IF)

should not perform as well as the intuitive-plus-thinking

or the sensing-plus-thinking types. Although IF types

perceive in a Gestalt manner and see the possibilities,

they evaluate on a subjective, good-versus-bad, personal

worth basis [26, pp. 54-55]. The cold formality of the

operational decision environment is not their forte.

Sensing-plus-feeling types are characterized by their

preference for hard facts in perceiving, just as the

sensing-plus-thinking types are [26, p. 54]. However, they

evaluate in the manner of the intuitive-plus-feeling type,

and are interested more in facts about people than about

things [26, p. 54]. Although subjects of this type should

not perform as well as those of the other types, it should

be emphasized that the measurement is relative only to

the performance of the other types, and is not absolute.

The implications of these results should be significant

in the area of MIS design. One implication is the IT and

ST type people make better production managers than IF and

SF types. There is some evidence to support this position

[26, p. 56]. If this is the case, then the Myers-Briggs

instrument could be used by firms as one tool in recruiting

production managers. IT and ST types would be ranked

above IF and SF types in the selection process.

The second hypothesis, H2, addresses the question of

how different combinations of report formats and levels of

summarization will affect user performance in general.







This question was explored by comparing the performance

measures of the subjects receiving the four different

combinations of report formats and levels of summarization:

Format Level of Summarization

1. Tabular (TAB) Raw Data (RD)
2. Tabular (TAB) Statistically Summarized Data (SSD)
3. Graphical (G) Raw Data (RD)
4. Graphical (G) Statistically Summarized Data (SSD)

Although there have been previous studies on the

effects of the level of summarization on performance D[, 6,

22, 30], the results have been mixed. The results of

studies on the effects of graphical versus tabular format

does indicate that higher performance is reached with the

graphic format in one [4] and with the tabular in two others

[1, 22].

It was anticipated that the highest performance would

come from the graphical-SSD report, followed by the graphi-

cal-raw data report, then the tabular-SSD report, and

finally the lowest performance from the tabular-raw data

report. Amador [1] and Lusk and Kersnick[22] showed higher

performance with tabular format, and Benbasat and Schroeder

[4] showed higher user performance using a graphical format.

Further, in one study, higher performance was demonstrated

with statistically summarized data [4], and in another

there was no significant difference between statistically

summarized and detail data [30]. Since the results of

these studies were not significant, and since there is some

evidence for the dominance of statistically summarized data,
the above ranking is suggested by prior research.








The implications in the area of MIS design are clear.

There are times when it would be impractical or even impos-

sible to consider the individual psychological types for

whom a report is intended. Therefore, the best report would

be the one which gave the highest performance over all

psychological types. This, of course, assumes that there

is no prior information concerning the proportion of dif-

ferent psychological types in the population of interest.

For example, it would be virtually impossible for a firm

to match all external and internal reports with the psy-

chological types of the users.

Perhaps the most significant question is raised by

the third hypothesis, H3, which proposes that there is one

best report combination for each psychological type of

user. It was anticipated that the IT type would perform

best with the G-SSD report combination, the ST type with

the TAB-RD combination, the IF type with the G-SSD combi-

nation, and the SF type with the TAB-RD. The reasons

for expecting these results are concerned with the charac-

teristics of each psychological type.

Since IT types prefer not to become embroiled with

large amounts of facts and since they evaluate in an

impersonal manner, it appears reasonable that they would

perform best with the graphical format and statistically

summarized data report. On the other hand, because of

the ST type's preference for detailed hard facts and im-

personal and logical evaluation of those facts, it also








seems reasonable to expect ST types to perform best with

the tabular format and raw data report.

The expectations for the IF and SF types were also

derived from the characteristics of the types, but were

arrived at in a different manner. Currently, MIS designs

do notpresent reports in what has been suggested as the

most effective form for those psychological types who

-use the feeling mode for evaluation. Mason and Mitroff

argue that "information for feeling types takes the form

of 'art,' 'poetry,' 'human drama,' and especially 'strong

moral component'" [23, p. 478]. Therefore, it was expected

that IF and SF psychological types would rely primarily on

their perception process. That is, IF types would perform

best with the G-SSD combination and SF types would perform

best with TAB-RD combination.

The implication of these findings for MIS design should

be highly significant. The implications are as follows:

1. A firm with an existing MIS report structure may
hire managers of the psychological type compatible
with that structure. A firm with a system that
produced G-SSD reports would hire IT types. If
the firm's system produced TAB-RD reports, then
ST types would be hired. A firm would not hire
SF or IF types in a production environment.

2. For a firm with an existing managerial complement,
top management could increase the performance of
the managers by using the Myers-Briggs instrument
to classify the managers by psychological type and
by designing the reports to accommodate that type
as follows:








Psychological
Type

IT
IF
-ST
SF


Report Format and
Level of Summarization


G-SSD
G-SSD
TAB-RD
TAB-RD


3. Even though the decision environment for this study
is limited to the production environment, an impli-
cation for the design of reports for top manage-
ment is also suggested. For reports containing
information for decisions similar to production
decisions, the report design should follow the
scheme for matching report design with psychologi-
cal type, as given in the preceding paragraph.













CHAPTER 3

EXPERIMENTAL RESULTS


Introduction

Results of the experiment pertaining to the hypotheses

are presented in this chapter. For each hypothesis, the

results from a MANOVA analysis of the simultaneous effect

on the three dependent variables if offered first. Then,

the results of a univariate analysis of variance for each

independent variable are presented, along with the results

from the application of Duncan's Multiple Range Test (DMRT)

for detecting significant differences between levels of

the appropriate independent variables. A result at p < .01

was considered highly significant, while p < .05 was

considered significant, and p < .10 was considered marginally

significant.

Results Pertaining to Hypothesis HI

The first hypothesis (HI) predicts that the psychologi-

cal type of the user will have a simultaneous effect on

the performance of the user as measured by the dependent

variables, total cost, decision time, and decision confi-

dence. The MANOVA result for analyzing the dependent

variables--time, confidence, and cost--by psychological

type demonstrates a significant simultaneous effect at

(F = 9.00, p < .027). Thus, the psychological type of





50

the decision maker does have a significant effect on the

three dependent variables jointly, and that portion of HI

is upheld.

The first hypothesis (H1) further predicts that sub-

jects of psychological type IT will have shorter decision

times, greater confidence in their decisions, and incur

less cost than subjects of the other three types. ST

subjects will outperform SF and IF subjects. IF subjects

will outperform SF subjects. The expected performance is

as follows:

IT > ST > IF > SF

where the > symbol represents "perform greater than".

To evaluate the individual differences between perfor-

mance by psychological types, a univariate analysis of

variance was run for each dependent variable, time, confi-

dence, and cost, by psychological type, i.e., ST, SF, IT,

IF, and the Duncan's Multiple Range Test (DMRT) for

detecting significant level differences. As shown in

Table 3, IT subjects had the lowest average decision time

(4.92 minutes), followed by SF subjects (5.88 minutes),

then IF subjects (6.08 minutes), and finally ST subjects

with the longest time (6.71 minutes). However, the only

significant difference was between IT and ST subjects.

Therefore the ordered relationship between time and psycho-

logical type predicted by the first hypothesis was not

supported.

Referring again to Table 3, ST subjects showed the

highest average confidence (7.75), followed in order by SF








and IF subjects at the same average confidence (7.71) and

IT subjects (7.46). The differences were not significant

and do not support that portion of the first hypothesis.

Results shown in Table 3 indicate that, costwise, SF

types performed better overall ($173,187.01) than the other

psychological types. Following SF types in cost performance

were ST types ($211,915.59), IF types ($242,763.95), and

IT types ($369,223.73) in that order. The univariate

analysis of variance of cost by psychological type sup-

ported the prediction that psychological type would affect

cost (F = 3.278, p < .0245). However, the differences

were only significant for SF versus IT types and ST versus

IT types. Therefore, that portion of H1 predicting signi-

ficant differences in performance by levels of psychologi-

cal types was not supported by the results of the experiment.

Even though a portion of H1 was not supported by the

experimental results, further analysis demonstrated a

significant difference between cost performance by subjects

based on perception (G). Sensing subjects obtained signi-

ficantly lower costs ($192,551.30) than intuitive subjects

($305,993.84), F = 5.735, p < .018. As well, univariate

analysis of cost performance differences obtained by

subjects based on the evaluation variable (E) showed

marginally significant lower costs ($207,975,48) for

thinking subjects than for feeling subjects ($290,569.66),

F = 2.955, p < .088.
















r- ui Cm 0

WC4-1 en' V n co
Cd wo C4 1-4 o 0
o 0 N r-C r-
0' 1-4 C%4 0n
11 0-1 r-.
0 01 o 'N 1
rj .


r-4




0


c0
4-

O


%. Q n L -1 4 -
r~- r- r^























* 0 )

e,
: 3 a r- *O c0

v4 -T 'o



















H 0H H V)
I-I C12 1-40 I


W
P4
>-I




0

C0
0




E-4 P


0)


0


0
SC
-) 0













0
>r(












*0

O
Co


l-








Results Pertaining to Hypothesis H2

The second hypohtesis (H2) concerns the combined

effect of format and the level of summarization on a

subject's performance. MANOVA comparisons indicate that

the type of the report used has a highly significant

effect on performance as measured simultaneously by time,

confidence, and cost (F = 3.00, p < .002). Further, as

predicted by the hypothesis, subjects with the different

report types should perform in the following order:

Highest: 1. Graphical Statistically Summarized Data
(G-SSD)

2. Graphical Raw Data (G-RD)

3. Tabular Statistically Summarized Data
(TAB-SSD)

4. Tabular Raw Data (TAB-RD)

Table 4 provides the results for average decision time,

average confidence, and average cost by report type.

According to the results shown in Table 4, subjects

with G-SSD report averaged the shortest decision time

(5.33 minutes) as predicted. However, the second shortest

decision time was obtained by subjects with the TAB-SSD

report (5.88 minutes) instead of those with G-RD report

(6.08 minutes) as predicted. Thus, the resulting order

of performance does not support the hypothesis for decision

time performance by report type. As well, the differences

observed for time by report type were not significant.








The predicted order of performance as measured by

confidence level was not substantiated by the results, as

shown in Table 4. Subjects receiving G-SSD reports finished

next to last in the rankings of confidence (7.33), while

subjects with G-RD reports or TAB-SSD reports exhibited

the highest confidence (8.08). However, even though the

predicted order of finish was not borne out, the differences

in confidence between subjects using G-RD and TAB-SSD and

those using G-SSD and TAB-RD are marginally significant.

A one-way analysis of variance of cost by report type

revealed a highly significant main effect on cost by report

type (F = 4.946, p < .003). However, subjects with G-SSD

reports were predicted to obtain the best cost performance.

Instead, as shown in Table 4, they obtained the highest

cost. G-RD reports provided the lowest cost, followed by

TAB-SSD, and finally by G-SSD. Therefore, the order of

performance by cost predicted by the hypothesis was not

borne out. But, the DMRT analysis demonstrated that cost

performance for subjects with G-RD and TAB-RD reports was

significantly better than those with TAB-SSD or G-SSD.

These results imply that, based on cost, subjects with RD

reports outperformed those with SSD reports. This impli-

cation was substantiated by the results of a one-way

analysis of variance of cost by the level of summarization

variable (L). Subjects using raw data (RD) reports obtained

a cost of $164,778.85 versus $333,766.29 for subjects using

statistically summarized data reports (SSD), (F = 13.75,


















C 0 0 0


(D < m %0
10 CO '
IT LI) CC 0 0

o o o0 m

o L C\ r-
Cm r-4 C4 r-I















00 00 v




















.4 (w o co o*
Q 0 &ccN








0OOH
ff *
*iimh

s^






C)OC C 0(Q
03 Q,
Qr \O Or
en Q
en o Q P







a i < <
0 0 E-i


PH

E-4

0


C.4
H
E-4

rt


<-4
3



ca


P3
cic





0













Ow
U


c)
< *M
(UC
QO


04








p < .0004). A highly significant simultaneous effect on

time, confidence, and cost by level of summarization (L)

was indicated by MANOVA results (F = 6.37, p < .0006).

The same MANOVA run showed no significant effect by format

(F).


Results Pertaining to Hypothesis H3

The third hypothesis (H3) proposes that there is a

best (in terms of performance) report type for each psycho-

logical type as follows:

Psychological "Best" Report
Type Type
IT G, SSD
ST TAB, RD
IF G, SSD
SF TAB, RD

Table 5 gives the results for IT types by report type,

while Table 6 provides the results for ST types. Results

for IF types are shown in Table 7, and those for SF types

are given in Table 8.

A MANOVA was run for each psychological type by report

type. The analysis showed a significant simultaneous

effect on time, confidence, and cost by report type when

the psychological type variable was controlled for all

psychological types, with the exception of IT's. For ST

types, the level of significance was p < .043, F = 2.171,

for IF types, p < .001, F = 3.785, and for SF types,

p < .00001, F = 8.073.








As shown in Table 5, IT types made decisions in the

least amount of time using G-RD reports (4.33 minutes).

However, they exhibited the greatest confidence with G-SSD

reports (8.50) and the least cost with TAB-RD reports

($221,375,50). Examination with DMRT showed no significant

differences for IT's by report in performance, measured in

time. IT's did show significantly less confidence in the

TAB-RD report type than in the other three report types

(p < .050), 5.83 for TAB-RD versus 7.67 for TAB-SSD, 7.83

for G-RD, and 8.05 for G-SSD. However, there was no signi-

ficant difference between IT's confidence in TAB-SSD, G-RD,

and G-SSD reports. Further, the cost differences by report

type for IT's were not significantly different. Although

the cost results were not significantly different, it is

interesting to note that IT types had the least confidence

(5.83, significant at p < .05) in the report type with

which they obtained the least cost (TAB-RD) and the

greatest confidence (8.50) in the report type with which

they obtained the highest cost (G-SSD).

As shown in Table 6, ST types took the least amount

of time for decisions using G-SSD reports (4.50 minutes),

had the greatest confidence in G-RD reports (8.00), and

obtained the least cost with G-RD reports ($93,302.56).

Univariate analysis by one-way ANOVA showed significant

main effects for time (F = 4.495, p < .0144), but not for

confidence or cost.



















1o a r- CM
i I, r- --T c14
1in -4 Cen -T
c- r- 'o0 r-4 -.
4-13W c 00 '00 0 o
S00
00 00 m
C4 m m- L















-l 0
0 C

Ca *H o00 r oo C









c-)r4




0
o E o m e E















o 090


>4
W
E-4
PQ








0



rP4
>4
M W














rt








Examination of time performance by ST's with DMRT

showed that ST's used marginally significant more time

to make decisions with TAB-SSD reports (9.67 minutes)

than with TAB-RD (6.83 minutes), G-RD (6.33 minutes),

and G-SSD (6.83 minutes). Also, ST's used significantly

less decision time with G-SSD reports than with TAB-RD

and TAB-SSD reports. However, the difference in time

involving G-SSD reports and G-RD reports was not signifi-

cant.

DMRT analysis of cost performance by ST types indicates

that ST's obtained marginally significant lower costs

($93,302.56) with G-RD types than with TAB-SSD ($238,113.26)

and G-SSD ($298,216.07) reports, but the difference between

G-RD and TAB-RD was not significant. The differences in

level of confidence were not significant.

As shown in Table 7, IF types exhibited the shortest

decision time with TAB-SSD reports (4.00 minutes), the

greatest confidence with TAB-SSD and G-RD reports (8.50)

and the least cost with G-RD reports ($79,239.98). Uni-

variate analysis by one-way ANOVA showed highly significant

effects for cost (F = 6.457, p < .003). The differences

for time and confidence were not significant.

Post-hoc DMRT analysis indicates that IF's used a

significantly larger amount of time to make a decision

using TAB-RD reports (8.33 minutes) than TAB-SSD reports

(4.00 minutes). The other time differences were not

significant. Nor were the differences in confidence among

the report types.






60














r-I O \ N LO
-T C Ln o

c 4J) C r-l 0 v-N
W (c0 0o r-4 c C1
0 0









P4

4-1 0

o o o co c~ c
Ei0 0(.
p r- Cm oC o
C r-q lt3 <0 00 C0 I>
CH c.


















m a a
w 44 w Cw











H
EEi

0



w 0 H A 0 q
0 PU
w a;




o 4 r cC 0
4) 1 1 O .








IF's using TAB-SSD reports had significantly higher

costs ($468,392.35) than IF's using any of the other report

types. The cost differences between IF's using TAB-RD,

G-RD, and G-SSD types were not significant.

In Table 8, one can see that SF types obtained the

shortest decision time using TAB-RD and TAB-SSD reports

(5.17 minutes). While SF's using TAB-SSD had the greatest

confidence (8.33), and those with the least cost had

TAB-RD reports ($83,650.62). Univariate analysis with

one-way ANOVA showed highly significant effects with cost

(F = 16.047, p < .00001) and time (F = 8.672, p < .007).

However, there was no significant effect on confidence.

Analysis by DMRT showed that SF's with TAB-RD and

TAB-SSD reports had significantly shorter decision times

(5.17 minutes) than SF's with G-RD (6.83 minutes) and

G-SSD (6.33 minutes). There was no significant difference

between decision times for SF's with G-RD and G-SSD reports.

SF's using TAB-RD reports ($83,650.62), TAB-SSD

reports ($83,973.88), and G-RD reports ($125,060.35) had

significantly lower costs than SF's using G-SSD reports

($400,063.21). There was no significant difference in

cost between SF's using TAB-RD, TAB-SSD, and G-RD reports.

DMRT analysis, as well, detected no significant differences

between the confidence of SF's using the different report

types.

The results concerning the third hypothesis (H3)

indicate that report type had a significant simultaneous

















o C n CO m
oo c'n (> so
r-4 N 4 N m V
4- Cj) aT Cr N C %

0 0 0- -z
r' o ccr CM









44. U
0U

*i ccn cn Cn


c,







SCo








0 U)
- (a c4J o co r

4JH*H ccl M i En
u *r oo * U 0











O Q C

)E- _4 <3 -43 1 l
:4 > E-4 0 U
a, < <
c n0


PLc
H









P4
>4


H P4


0


E-4



My




















0O CO) 0 C
4 to o in ( o Co
00 'O 0 0 0 0

00 0 CC O









V). V). V).


















O c
r o 0 oo










00





4 C)



S*o






0 a ;)






o 00 l

cU E-4 < < O
O94 E-4 -


PH




0
P4





00 CE
W
W P0






Fr
E-4



H


C1





64

effect on all three dependent variables (time, confidence,

and cost) for all psychological types with the exception

of the IT type. However, no report type was significantly

better for IT types than any other. Also, there was no

one best report type for ST, SF, or IF types. Therefore,

the third hypothesis (H3) was not supported by experimental

results.














CHAPTER 4

DISCUSSION OF RESULTS


Introduction

The experimental results are discussed in the following

pages with respect to the research implications stated in

Chapter 2. In addition, the experimental results presented

in Chapter 3 are.compared to the results of previous

studies. Finally, the results of the post-experimental

questionnaire are presented and their relationship with

the experimental results are discussed.


Implications of Hypothesis H1

The implications following from the first hypothesis

are concerned with the selection of employees as production

managers if these experimental results could be demonstrated

in an actual production environment. Assuming that mini-

mization of production costs is desirable for an organi-

zation, then the organization should use, as one of its

selection criteria, the decision-making type of the prospec-

tive employee.

Since SF and ST types performed significantly better

than IT types in cost performance, given a choice among

SF, ST, and IT types with all other things being equal,

either the ST or SF types should be selected. Further, if








the organization has information on only the perceptual

type of the prospective employees available for use in the

selection process, then the sensing type should be selected,

all other things being equal. This inference is indicated

by the significantly better cost performance of sensing

subjects versus intuitive subjects reported in Chapter 3.

If, on the other hand, the only information concerning

the decision-making type of the individual is their evalua-

tion type, then the thinking type should be selected, all

other things being equal. This inference is made from the

significantly better cost performance of thinking subjects

versus that of feeling subjects reported in Chapter 3.

Implications of Hypothesis H2

The implications of hypothesis H2 as outlined in

Chapter 2 addresses the situation in which an organization

lacks knowledge of the distribution of decision-making

types of individuals receiving reports within the organi-

zation. Another situation an organization faces is the

design of reports for external users. In both of the

foregoing situations, or any others where it would be

impractical or impossible to consider the psychological

type of each individual for whom a report is intended,

the best report would be the one which provided the best

performance over all psychological types. Given that the

experimental results were replicated in an actual production

environment.









Results reported in Chapter 3 pertaining to hypothesis

H2 show that, costwise, subjects using G-RD and TAB-RD

reports significantly outperformed subjects using G-SSD

and TAB-SSD reports, with no significant difference ob-

served between subjects with G-RD and TAB-RD reports or

between subjects with G-SSD and TAB-SSD reports. Further

investigation demonstrated that subjects using RD reports

performed significantly better costwise than subjects

using SSD reports, while differences in format were not

significant.

Although the results are not generalizable across all

decision environments, they do suggest that for reports

within a production type of environment, the best report

for cost performance where the decision type of the user

is unknown is one with raw data (RD) level of summarization

either tabular (TAB) or graphical (G) format. This suggests

that MIS designers, for the most part, have been doing

the right thing in producing primarily TAB-RD or G-RD

reports.


Implications of Hypothesis H3

The first implication following from the results for

hypothesis H3 as outlined in Chapter 2 is concerned with

the organization which has an existing MIS report structure.

Assuming these experimental results could be found in an

actual production situation,it is logical that the firm

would be interested in hiring managers whose psychological








type is compatible with that structure. Following the

above logic, a firm with an existing report structure

dominated by TAB-RD reports should hire SF or IF types

as production managers. This: inference follows from

the experimental results which showed SF and IF types

performing significantly better costwise with TAB-RD

reports.

For a firm whose existing MIS structure is dominated

by G-RD reports, the above logic dictates the hiring of

ST, IF, or SF types as production managers. Again, the

inference which follows from experimental results shows

that SF types performed better with G-RD reports than with

TAB-RD and G-SSD reports. Further, SF and IF types showed

significantly better cost performance with G-RD reports

as well as with TAB-RD reports, as mentioned in the

previous paragraph, and therefore would qualify for

priority consideration as production managers in G-RD

report-dominated MIS systems, along with ST types.

For a firm with G-SSD report-dominated MIS systems,

there appears to be only one decision-making type demon-

strating compatibility with this report structure. Subjects

of the IF type were the only ones whose cost performance

with G-SSD reports was not worse than their cost perfor-

mance with the other three reports. Further, the dif-

ferences in cost performance was significant for ST's be-

tween G-RD and G-SSD reports, and for SF's between G-SSD

and all other reports types.








A firm whose MIS is dominated by a TAB-SSD report

structure should consider SF types for production managers,

all other things being equal. Experimental results indi-

cate that SF types demonstrated significantly lower costs

with TAB-SSD, G-RD, and TAB-RD reports versus G-SSD reports,

and thus are compatible with the TAB-SSD reports structure.

Even though IT types showed no significant difference

in cost performance with different report types, their

poor cost performance relative to the other decision-

making types indicates their incompatibility with the

production environment itself. Perhaps, the implication

is that IT types should not be considered for production

employment, all other things being equal.

Another implication from hypothesis H3 is that, for

a firm with an existing managerial complement, performance

for the firm's production managers may be increased using

the Meyers-Briggs instrument to classify production managers

by psychological type and designing the production reports
to accommodate the managers' psychological type. Experi-

mental results indicate that managers may be able to use

more than one report type, as explained in the following

paragraphs.
Based on cost performance alone, production managers

of the ST type should receive G-RD reports, even though

they use less time with both G-SSD and G-RD reports.

Results of the experiment indicate ST's do have the

greatest confidence in G-RD reports; however, the difference

was not significant.








SF production managers can use either TAB-RD, TAB-SSD,

or G-RD reports, according to experimental cost performance

results. Since differences in the costs between SF subjects

using those reports can be due to random effects, either of

the three report types would be appropriate for SF's. How-

ever, SF production managers should use less decision time

with TAB-RD and TAB-SSD reports than G-RD reports. Thus,

the findings suggest that SF production managers would

benefit from TAB-RD or TAB-SSD reports.

Experimental cost performance by IF subjects suggests

that IF production managers should receive either TAB-RD,

G-RD, or G-SSD reports, since there was no significant

difference between the performance of subjects using those

reports, and the difference between cost performance of

subjects with those reports and subjects with TAB-SSD

reports were significant, as reported in Chapter 3. How-

ever, since IF subjects with G-RD reports used significantly

more decision time than those using TAB-SSD reports, G-RD

reports show conflicting performance results and are

eliminated from suggested reports for IF production managers.

Experimental results show no significant differences in

confidence for IF subjects. Thus the findings imply that

IF production managers would benefit from TAB-RD and G-SSD

reports.

As mentioned previously, the relatively poor perfor-

mance of IT types in this experiment suggest that the

production environment is not the forte of IT types.








Further, based on cost performance, no report type is

better than any other for IT types, since the differences

between report types were not significant for IT's, as

reported in Chapter 3. Decision time was not a differen-

tiating factor either. However, experimental results

did show significantly less confidence in TAB-RD report

type than the other report types. This finding does

suggest, rather weakly, that the TAB-RD report could be

eliminated from those presented to IT production managers,

based on confidence.

The final implication arising from hypothesis H3, as

outlined in Chapter 2, is concerned with the design of

reports for top management. Even though the decision

environment for this study was limited to production, and

the findings are not generalizable beyond that environment,

it is suggested that, for top management reports containing

information similar to production decisions, the report

design should follow the scheme for report designs consis-

tent with the findings of this study. Thus, top managers

of a particular psychological type would receive the report

type specified in the previous paragraphs.

Top managers of the ST type would receive information

for production-type decisions in G-RD reports, SF top

managers either TAB-RD or TAB-SSD reports, IF top managers

either TAB-RD or G-SSD reports, while IT top managers may

use any of the report types with relatively the same results.








Comparison with Other Studies

The results of the present study concerning the effects

of the DM and CIS variables confirm the results of some

previous studies and conflict with others. The following

paragraphs describe the confirmation and differences, and

suggest some reasons for the differences.

Previous studies of the effects of cognitive style (DM)

on performance within the context of an MIS by Benbasat and

Schroeder [4], Kozar [20], and Lusk and Kersnick [22]

have shown no significant effects due to the cognitive type

of the decision maker. The primary difference between the

structure of the studies pertaining to cognitive type is

the instrument used to measure the attribute as detailed

in Chapter 1. Benbasat and Schroeder used the WCFAFT [4],

Kozar used quantitative scores on the ATGSB [20], Lusk and

Kersnick used the EFT while the MBTI was used in the present

study.

As reported in Chapter 3, significant effects, both

univariate and multivariate, were found due to cognitive

type of the decision maker. The foregoing suggests that

the MBTI is the instrument that will provide the most

rewarding results in future MIS research on cognitive style.

As presented in Chapter 1, results from previous

studies have been mixed on the effects of format on user's

performance. Amador's results [1] and Lusk and Kersnick

[22] showed higher performance with tabular format, while

Benbasat and Schroeder's results showed higher performance









using graphical format [4]. The present study, as stated

in Chapter 4, found no significant differences in performance

due to main effects of format using either univariate or

multivariate models of analysis of variance for investiga-

tion. A possible reason for the lack of a significant main

effect due to format is the similarity of the tabular and

graphical formats (Appendix D). The use of bar graphs to

represent graphical form might not offer sufficient differ-

entiation of graphical versus tabular format.

The effect of the level of summarization variable on

performance found in previous studies is mixed, as reported

in Chapter 1. Amador [1] found significant results using

a multivariate model, while Chervany and Dickson [6], Senn

and Dickson [30], and Lusk and Kersnick [22], using univar-

iate models for investigation, did not. The present study,

as reported in Chapter 4, found that subjects with raw data

(RD) reports significantly outperformed subjects with statis-

tically summarized data (SSD) reports, when analyzed with

either univariate or multivariate models.


Discussion of Post-Experimental Questionnaire Results

Results of the post-experimental questionnaire are shown

in Table 9. First, a distribution of report type preferences

is shown by psychological type. Next is shown the distribu-

tion by psychological type of the percent of that type which

preferred a report type other than the one received in the

study. The last distribution shows the percent of subjects

by psychological type that wanted additional information.








As indicated in the first distribution, 87 percent of

ST's preferred TAB-RD reports, while 13 percent preferred

TAB-SSD reports, with none preferring the other two report

types. Further, examination of Table 9 shows that 63 percent

of the ST's preferred a report other than the one they

received. This preference for the TAB-RD report by ST's

conflicts with the experimental results, which show the

best performance costwise by ST's is with G-RD reports

versus TAB-SSD and G-SSD reports, and no significant dif-

ference between performance by ST's with TAB-RD reports

and other report types. The foregoing suggests that the

report types preferred by ST production managers are

not the ones they use to the best advantage.

Table 9 shows that only 25 percent of ST's wanted *

additional information in their reports, while 75 percent

were satisfied with the information the recieved. These

figures indicate that ST subjects were satisfied with the

content of the reports supplied for the experiment.

Table 9 shows that 87 percent of SF's preferred TAB-RD

reports, while 13 percent preferred TAB-SSD reports--the

same distribution demonstrated by ST's. Further, 50 percent

of the SF's preferred a different report than the one they

received. The preferences demonstrated by SF's is not in

conflict with experimental results, since SF's cost perfor-

mance with TAB-RD reports was the best, although not








significantly different from performance with TAB-SSD

and G-RD reports.

SF subjects viewed additional information the same as

ST types did. Only 25 percent of SF's wanted additional

information. Again, this suggests that SF's, for the

most part, were satisfied with the content of the reports

they received during the experiment.

Table 9 shows that 61 percent of IT's preferred TAB-RD

reports, 6 percent preferred G-RD reports, 17 percent

preferred G-SSD reports, percent preferred TAB-SSD

reports, and 10 percent preferred a report type other

than the ones offered in the experiment. Further, 83

percent of IT's reported a preference for a report type

other than the ones they received. The reported prefer-

rences by IT's are not inconsistent with the experimental

results. IT subjects performed best costwise with TAB-RD

report type. However, there was no significant difference

between the performance of IT's using the TAB-RD and

IT's using other report types.

Only 16 percent of the IT subjects reported a desire

for additional information. Thus IT's were, for the most

part, satisfied with the content of the reports received

in the experiment.

Report preference by IF's from Table 9 indicate that

75 percent of the IF's preferred TAB-RD reports, while 13

percent preferred G-RD reports and 12 percent preferred

G-SSD reports. Also shown in Table 9 is the fact that








63 percent of IF's were unhappy with the report types they

received and would have preferred another type of report.

The reported preferences are consistent with the experimen-

tal results for IF performance. Costwise, IF subjects

performed significantly worse with TAB-SSD reports than IF's

using any other report type. As reported in Chapter 4, IF's

performed best with G-RD reports, followed by TAB-RD reports

and G-SSD reports, in that order. However, the differences

between these three reports were not significant.

As shown in Table 9, IF subjects were the only psycho-

logical category to want additional information. They

indicated that content of the reports received was not

sufficient by 63 to 27 percent.

Overall, subject responses were not inconsistent with

their experimental performance. The overwhelming preference

for TAB-RD reports could be a matter of conditioning due

to the prevalence of TAB-RD reports incorporated in most

information systems report structures. Also, the dis-

satisfaction of the majority of subjects with the report

type they received implies a lack of familiarity with the

more unorthodox report types, not necessarily their inability

to perform with those report types. Finally, the dissatis-

faction reported by IF types with the content of the reports

received in the study could well be due to an attribute of

that psychological type suggested earlier in Chapter 1,

that IF's perform best and are more attuned to narratives

and visual imagery.













TABLE 9

RESULTS OF POST-EXPERIMENTAL QUESTIONNAIRE


Report Preferences (Percent)


ST SF IT IF

87 87 61 75

0 0 6 13

0 0 17 12

13 13 6 0

0 0 10 0



Percent That Preferred Report
Other Than One Received

ST SF IT IF

63 50 83 63


Percent That Wanted


Additional Data


ST SF IT IF

25 25 16 63


TAB-RD

G-RD

G-SSD

TAB-SSD

OTHER













CHAPTER 5

SUMMARY AND POSSIBLE EXTENSIONS


This research was motivated by the relative scarcity

of empirical research dealing with management information

systems and the lack of significant findings and conflicting

evidence from previous studies pertaining to the effects of

cognitive type of the user and report type on user perfor-

mance. An experimental model for user performance involving

three dependent variables (user confidence, decision time,

and production cost) and three independent variables

(characteristics of the decision maker, characteristics of

the decision environment, and characteristics of the infor-

mation system) was used in the study. The model was used

to examine the effects on performance by four different

report types and four different psychological (decision

making) types of production managers. Ninety-six MBA

students participated as surrogate managers.

Three general hypotheses were tested relative to the

effects of the report type (format and level of summariza-

tion) treatments on the performance of subjects of each

psychological type. Portions of each hypothesis were

supported by the experimental results. However, as with

all experimental research, questions concerning the results









of this study are possible, and in that area, the present

study is not exempt.

As pointed out by Amador [1, p. 63], using students as

surrogate managers is a limitation in any study. A neces-

sary extension of the present study would be an experimental

study using real-life production managers. A further exten-

sion would be controlled studies with real-life production

managers in a real-world site. This would provide external

validity for the study and important additions to knowledge

in the MIS design field.

Further investigation is needed into the relatively

poor performance of IT types in the production environment.

Perhaps this particular psychological type is more attuned

to other decision environments in an organization. A modi-

fication of the experiment to consider other decision en-

vironments is a logical extension.

Another area requiring study is the apparent dissatis-

faction of IF types with the traditional forms of reports,

as indicated by their responses on the post-experimental

questionnaire. As suggested by Mason and Mitroff [23],

IF's may perform better with narrative reports and pictures.

In addition to increasing the types of reports, a

greater differentiation is required between tabular and

graphic report formats. The lack of a significant effect

due to format is very possibly due to the close similarity

of the representations of tabular and graphic in the present









study. Possibly the use of continuous plots to represent

graphical format would be the key.

Another area of consideration is the extension of the

decision environment to consider strategic level decisions.

Items of interest would be what kinds of report types

benefit performance at that level, and what psychological

types provide the best performance at that level.

Finally, the results of this study imply that, as

suggested by Amador [1] and Barkin and Dickson [3], the use

of multivariate models are of significant value in this type

of study. Where effects on dependent variables considered

individually are sometimes not detected, those effects on

all the dependent variables considered jointly can be sig-

nificant. In particular, each MANOVA test in the present

study detected significant results, while many of the uni-

variate tests did not. The use of multivariate tests permits

the use of information lost in the univariate tests.















REFERENCES


1. Amador, Jose A., "Information Formats and Decisions
Performance; An Experimental Investigation",
Ph.D. Dissertation, University of Florida, 1977.

2. Aron, Joel D., "Information Systems in Perspective",
Computing Surveys, Vol. 1 (December, 1969),
pp. 213-236.

3. Barkin, S.R., and Dickson, G.W., "An Investigation of
Information System Utilization", Information and
Management, Vol. 1 (November, 1977), pp. 35-45.

4. Benbasat, I., and Schroeder, R.G., "An Experimental
Investigation of Some MIS Design Variables", The
Management Information Systems Quarterly, Vol. 1
(March, 1977), pp. 37-49.

5. Carlyn, Marcia, "An Assessment of the Myers-Briggs
Type Indicator", Journal of Personality Assessment,
Vol. 45, No. 5 (1977), pp. 461-473.

6. Chervany, N.L., and Dickson, G.W., "An Experimental
Evaluation of Information Overload in a Production
Environment", Management Science, Vol. 20 (June,
1974), pp. 1335-1344.

7. Churchill, Gilbert A., Jr., Marketing Research,
Methodological Foundations, 2nd ed., The Dryden
Press, Hinsdale, Illinois, 1978.

8. Davis, Gordon, B., Management Information Systems:
Conceptual Foundations, Structure and Development,
McGraw-Hill, New York, 1974.

9. Dickson, G.W., Chervany, N.L., and Kozar, K.A., "An
Experimental Gaming Framework for Investigating
the Influence of Management Information Systems
on Decision Effectiveness", Management Information
Systems Working Paper Series, MISRC-WP-72-12,
University of Minnesota, Minneapolis, Minnesota,
April, 1972.









10. Senn, J.A., and Chervany, N.L., "Research
in Management Information Systems: The Minnesota
Experiments", Management Science, Vol. 23 (May, 1977),
pp. 913-933.

11. Driver, M.J., and Mock, T.J., "Human Information
Processing, Decision Style Theory, and Accounting
Information Systems", The Accounting Review, Vol. 1
(July, 1975), pp. 49'0-508.

12. Fellingham, J.C., Mock, T.J., and Vasarhelyi, M.A.,
"Simulation of Information Choice", Decision
Sciences, Vol. 7 (April, 1976), pp. 219-234.

13. Fiedler, Fred E., "A Contingency Model of Leadership
Effectiveness", Advances in Experimental Social
Psychology, Berkowitz, Leonard, ed., Academic
Press, New York, 1964, Vol. 1, pp. 149-190.

14. Gibbons, Jean D., Nonparametric Methods for Quantitative
Analysis, Holt, Rinehart, and Winston, New York,
1976.

15. Hanold, Terrance, "An Executive View of MIS",
Datamation, Vol. 11 (November, 1972), pp. 65-71.

16. Harris, Roy D., and Maggard, Michael J., Computer
Models in Operations Management, Harper and Row
Publishers, Inc., New York, 1972, pp. 197-220.

17. Hellriegel, Don, and Slocum, J.W., Jr., "Managerial
Problem Solving Styles", Business Horizons, Vol.
18, No. 6 (December, 1975), pp. 27-39.

18. Jung, C.G., Analytical Psychology, Its Theory and
Practice, Vintage Books, New York, 1968.

19. Kilmann, Ralph H., and Taylor, Vern, "A Contingency
Approach to Laboratory Learning: Psychological
Types Versus Experimental Norms", Human Relations,
Vol. 27, No. 9 (1974), pp. 891-909.

20. Kozar, K.A., "Decision Making in a Simulated Environ-
ment: A Comparative Analysis of Computer Display
Media", Ph.D. Dissertation, University of Minne-
sota, 1973.

21. Lucas, H.C., "Performance and the Use of an Information
System", Management Science, Vol, 21 (April, 1975),
pp. 908-919.

22. Lusk, Edward J., and Kersnick, Michael, "The Effect of
Cognitive Style and Report Format on Task Performance:
The MIS Design Consequences", Management Science,
Vol. 29 (August, 1979), pp. 787-798.









23, Mason, R.D,, and Mitroff, ,I., "A Program for Research
on Management Information Systems", Management
Science, Vol, 19 (January, 1979), pp. 475-547,

24. McKenney, James L., and Keen, P.G., "How Managers'
Minds Work", Harvard Business Review, Vol. 44,
No. 3 (May-June, 1974), pp. 79-90.

25. Mitroff, II., and Kilmann, R.H., "On the Importance of
Qualitative Analysis in Management Science: The
Influence of Personality Variables on Organizational
Decision-Making", Working Paper, Graduate School of
Business, University of Pittsburgh, Pittsburgh,
Pennsylvania, 1974.

26. Myers, I.B., Myers-Briggs Type Indicator, Educational
Testing Service, Princeton, New Jersey, 1962.

27. Pokempner, Stanley J., "Management Information Systems--
A Pragmatic Survey", The Conference Board Record
Vol. 1 (May, 1973), pp. 49-54.

28. Radford, K.J., Information Systems for Strategic
Decisions, Prentice-Hall, Inc., New York, 1978.

29. Schroeder, R.G., and Benbasat, I., "An Experimental
Evaluation of the Relationship and Uncertainty in
the Environment to Information Used by Decision
Makers", Decision Sciences, Vol. 6 (July, 1975),
pp. 556-567.

30. Senn, J.A., and Dickson, G.W., "Information System
Structure and Purchasing Decision Effectiveness",
Journal of Purchasing and Materials Management,
Vol. 10 (August, 1974), pp. 776-783.

31. Shaw, Marvin E., Scaling Group Tasks: A Method for
Dimensional Analysis, Technical Report No. 1,
University of Florida, July, 1963.

32. Simon, Herbert A., The New Science of Management
Decision, Prentice-Hall, Inc., New York, 1977.

33. Small, John T., and Lee, William B., "In Search of MIS",
MSU Business Topics, Vol. 1 (Autumn, 1975), pp. 47-55.

34. Smith, August., W., and Urban, Thomas F., "Myers-Briggs
Personality Orientations and Information Processing
Styles; Implications for Management Education and
Development", National Academy of Management,
MED Session, San Francisco, August 10, 1978.






84


35. Van Gigch, John P., Applied General Systems Theory,
2nd ed., Harper and Row, New York, 1978.

36. Witkin, H., Dyk, R., Faterson, H., Goodenough, D.,
and Karp, S., Psychological Differentiation,
Wiley and Sons, New York, 1962.

37. Goodenough, D., and Karp, S., "Stability
of Cognitive Style from Childhood to Young Adult-
hood", Journal of Personality and Social Psychology,
Part I, Vol. 7 (November, 1967), pp. 73-84.

38. Zmud, Robert W., "On the Validity of the Analytic-
Heuristic Instrument Utilized in The Minnesota
Experiments", Management Science, Vol. 24 (June,
1978), pp. 1088-1098.






























APPENDIX A

SIMPRO LISTING








PROGRAM SIMPRO(OUT,MASIN1,MASIN2=/225,WORKIN,WORKOT,DECS,
+ MSOUT1,MSOUT2=/225,OUTPUT,TAPE6=OUTPUT,
+ TAPE4=OUT,TAPE7=MASIN1,TAPE8=MASOUT1,TAPE9=WORKIN,
+ TAPE10=WORKOT,TAPE11=DECS,TAPE12=MASIN2,TAPE14=MSOUT2)
COMMON PRPL(50),DMDNR(50),DMD(50),CPCG,CEXI,CSKO,
+ CRI,ISN,INM1,INM2,ITP,IBLK,RSTA(10,3),NBR,BI,TPFC,TEIHC,
+ TRIHC,TSKOC,TCOST,OSPRO,BO,ENIN(50),PRFL(50),
+ RIC(50),EIC(50),SOC(50),TCW(50),EI,PDCOST
CPCG= 0.10
CRI= 0.02
CEXI = 0.06
CSKO = 0.25
REWIND 7
REWIND 9
REWIND 11
XM1=SECOND(CP)
XM1=100*XM1
NN=INT(XM1)
JJ=MOD(NN,2)
IF(JJ .EQ. 0) GO TO 5
M1=NN
GO TO 6
5 M1=NN+1
6 M2=0
20 READ(9,2100) NIAW,IBLKW,NBR,BI,TPFC,TRIHC,TEIHC,TSKOC,
+ TCOST,BO,OSPRO
IF (NBR.GT. 50) CALL ATEND(TCOST)
30 READ(7,2000) ISN,INM1,INM2,ITP,IBLK
READ(12,2010) ((RSTA(I,J),J=1,3),I=1,10)
N = NBR + 4
READ(11,2200) NID,DT,DC,(PRPL(I),I=NBR,N)
CALL FCAST(NBR)
CALL ADMAND(NBR,M1,M2,RN)
CALL PROD
J = NBR + 5
-CALL FCAST(J)
CALL POUT
NBR = NBR + 5
CALL UPDT(NIAW,IBLKW,DT,DC)
IF (NBR .GT. 50) CALL ATEND(TCOST)
2000 FORMAT(I9,2A10,A4,I2)
2010 FORMAT(8(F12.2,F4.0,F3.0)
2100 FORMAT(19,2I2,F8.0,5F10.2,2F8.0)
2200 FORMAT(I9,F4.0,5F8.0)
END
SUBROUTINE FCAST(NWK)
COMMON PRPL(50),DMDNR(50)
N=NWK=4
10 DO 20 I = NWK,N
20 DMDNR(I)=100000+(30000*SIN((I/12.0)*6.28318))+1000*I
RETURN
END
SUBROUTINE ADMAND(NBR,M1,M2,RN)
COMMON PRPL(50),DMDNR(50),DMD(50)






87



M = NBR+$
DO 40 I = NBR, M
TM = DMDNR(I)
TSD = .05*TM
Z = 0
DO 30 N = 1,12
CALL RAN(M1,M2,RN)
Y=RN
30 Z = Z + Y
RV= (Z-6)/4
40 DMD(I) = RV TSD + TM
RETURN
END
SUBROUTINE PROD
COMMON PRPL(50),DMDNR(50),DMD(50),CPCG,CEXI,CSKO
+ CRI,ISN,INM1,INM2,ITP,IBLK,RSTA(10,3)NBR,BI,TPFC,TEIHC,
+ TRIHC,TSKOC,TCOST,OSPRO,BO,ENIN(50),PRFL(50),
+ RIC(50),EIC(50),SOC(50),TCW(50),EI,PDCOST
M=NBR+4
PDCOST=0.
IF (NBR .EQ. 1) OSPRO = PRPL(1)
DO 32 IT = NBR, M
PROSC = PRPL(IT)
El = BI+PROSC-DMD(IT)-BO
IF (El) 26,27,27
26 BO = -El
El = 0
GO TO 28
27 BO = 0
28 CONTINUE
PFC = ABS(OSPRO-PROSC)*CPCG
IF (EI-240000) 29,29,30
29 RIHC = EI*CRI
EIHC = 0
GO TO 31
30 EIHC = (EI-240000)*CEXI
RIHC = 240000*CRI
31 CONTINUE
SKOC = BO*CSKO
TOTC = PFC+RIHC+EIHC+SKOC
ENIN(IT) = El
PRFL(IT) = PFC
RIC(IT) = RIHC
EIC(IT) = EIHC
SOC(IT) = SKOC
TCS(IT) = TOTC
PDCOST=PDCOST + TOTC
TPFC = TPFC + PFC
TRIHC = TRIHC + RIHC
TEIHC = TEIHC + EIHC
TSKOC = TSKOC + SKOC
TCOST = TCOST + TOTC
BI = El
OSPRO = PROSC










32 CONTINUE
RETURN
END
SUBROUTINE POUT
COMMON PRPL(50),DMDNR(50),DMD(50),CPCG,CEXI,CSKO,
+ CRI,ISN,INM1,INM2,ITP,IBLK,RSTA(10,3),NBR,BI,TPFC,TEIHC,
+ TRIHC,TSKOC,TCOST,OSPRO,BO,ENIN(50),PRFL(50),
+ RIC(50),EIC(50),SOC(0)),TCW(50),EI,PDCOST,XAD,XPRD,
+ XENI,XPRFL,XRIC,XEIC,XSOC,XTCW,XDMD,SAD,SPRD,
+ SENI,SPRFL,SEIC,SSOC,STCW,SDMD
WRITE(4,900)
WRITE(4,1000)
N = NBR + 4
WRITE(4,1010) NBR,N,INM1,INM2,ISN
WRITE(4,1020)
SAD=SPRD=SENI=SPRFL=SRIC=SEIC=SSOC=STCW=SDMD=0
XAD=XPRD=XENI=XPRFL=XRIC=XEIC=XSOC=XTCW=XDMD=0
DO 12 K=NBR,N
XAD=XAD+DMD (K)
RD=XP=XPR+PRPL (K)
XENI=XENI+ENIN(K)
XPRFL=XPRFL+PRFL(K)
XRIC=XRIC+RIC(K)
XEIC=XEIC+EIC(K)
XSOC=XSOC+SOC(K)
XTCW=XTCW+TCW (K)
II=K+5
12 XDMD=XDMD+DMDNR(II)
XAD=XAD/5
XPRD=XPRD/5
XENI=XENI/5
XPRFL=XPRFL/5
XRIC=XRIC/5
XEIC=XEIC/5
XSOC=XSOC/5
XTCW=XTCW/5
XDMD=XDMD/5
DO 13 K=NBR,N
SAD= SAD + (DMD(K) XAD)**2
SPRD = SPRD + (PRPL(K) XPRD)**2
SENI = SENI + (ENIN(K) XENI)**2
SPRFL = SPRFL + (PRFL(K) XPRFL)**2
SRIC = SRIC + (RIC(K) XRIC)**2
SEIC = SEIC + (EIC(K) XEIC)**2
SSOC = SSOC + (SOC(K) XSOC)**2
STCW = STCW + (TCW(K) XTCW)**2
II=K+5
13 SDMD = SDMD + (DMDNR(II) = XDMD)**2
SAD=SQRT(SAD/5)
SPRD=SQRT(SPRD/5)
SENI=SQRT(SENI/5)
SPRFL=SQRT(SPRFL/5)
SRIC=SQRT(SRIC/5)









SEIC=SQRT(SEIC/5)
SSOC=SQRT(SSOC/5)
STCW=SQRT(STCW/5)
SDMD=SQRT(SDMD/5)
IF((IBLK .GE. 13) .AND. (IBLK L#. 16)) CALL GSD(PDCOST,
+XAD,XPRD,XENI,XPRFL,XRIC,XEIC,XSOC,XTCW,XDMD,SAD,SPRD,SENI,
+ SPRFL,SRIC,SEIC,SSOC,STCW,SDMD)
IF((IBLK .GE.5) .AND. (IBLK .LE. 8)) CALL TABSD(PDCOST,
+ XAD,XPRD,XENI,XPRFL,XRIC,XEIC,XSOC,XTCW,XDMD,SAD,SPRD,SENI,
+ SPRFL,SRIC,SEIC, SSOC,STCW,SDMD)
IF((IBLK .GE. 9) .AND. (IBLK .LE.12)) CALL GRD(NBR,ENIN,
+ PRFL,RIC,EIC,SOC,TCW,PRPL,PDCOST,DMD,DMDNR,IBLK)
IF((IBLK .GE. 1) .AND. (IBLK .LE 4)) CALL TABRD(NBR,ENIN,
+ PRFL, RIC, EIC, SOC, TOCW, PROPRL,PDCOST,DMD,DMDNR,IBLK)
900 FORMAT(1H1,///)
1000 FORMAT(57X,*OPERATING RESULTS*,/)
1010 FORMAT(1HO,45X,*FOR WEEKS *,12,* *,12,* MANAGER *,
+ 2A10,X,I9,//)
1020 FORMAT(1H0,13X,*- UNITS -- *
+ *- -*,8X,*- - COSTS*
RETURN
END
SUBROUTINE TABRD(NBR,ENIN,PRFL,RIC,EIC,SOC,TCW,
+ PRPL,PRDCOST,DMD,DMDNR)
DIMENSION ENIN(50),PRFL(50),RIC(50),EIC(50),SOC(50),
+ TCS(50),PRPL(50),DMD(50),DMDNR(50)
WRITE(4,1030)
WRITE(4,1040)
N = NBR + 4
DO 20 I = NBR, N
20 WRITE(4,1050) I,DMD(I),PRPL(I),ENIN(I),PRFL(I)
+ RIC(I),EIC(I),SOC(I),TCW(I)
WRITE(4,1060) PDCOST
WRITE(4,1070)
WRITE(4,1080)
WRITE(4,1090)
J = NBR + 9
N = NBR + 5
DO 40 I = N,J
40 WRITE(4,1100) I,DMDNR(I)
1030 FORMAT(1H0,25X,*ACTUAL*,5X,*NUMBER*, 7X,*ENDING*,7X,
+ *PRODUCTION*,6X,*REGULAR*,8X,*EXTRA*,11X,*STOCK-*,
+ 6X,*TOTAL WEEKS*)
1040 FORMAT(1H ,14X,*WEEK*,7X,*DEMAND*,5X,*PRODUCED*,4X,
+ *INVENTORY*,6X,*FLUCUATION*,5X,*INVENTORY*,5X,
*INVENTORY*,10X,*OUTS*,10X,*COST*//)
1050 FORMAT(1H ,15X,I2,5X,F11.0,3X,F8.0,4X,F8.0,6X,F11.2,
+ 3X,F11.2,3X,F11.2,3X,F11.2,5X,F12.2,/)
1060 FORMAT(1H0,93X,*TOTAL COST FOR THIS PERIOD*,F13.2,////)
1070 FORMAT(1H0,10X,10(3H- -))
1080 FORMAT(1HO,48X,*DEMAND FORECAST FOR NEXT FIVE WEEKS*)
1090 FORMAT(1H0,14X,*WEEK*,7X,*DEMAND*,//)
1100 FORMAT(1H ,15X,I2,7X,F8.0,/)
RETURN











END
SUBROUTINE FIXARY(NBR,XRAY,IARY,ARY)
DIMENSION XRAY(50),IARY(10,5),ARY(10),TEMP(10),TEN(10)
DO 5 I=1,10
ARY(I) = 0.
DO 4 J=1,5
4 IARY(I,J) = 1H
5 CONTINUE
N= NBR+4
J = 0
j-0
DO 10 I= NBR,N
J= J+1
10 TEMP(J) = XRAY(I)
JJ=0
DO 15 I = 1,10
IF (TEMP(I) .NE. 0.) JJ=1
15 CONTINUE
IF (JJ .EQ. 0) GO TO 80
COMP= TEMP(1)
DO 20 I= 2.5
IF (TEMP(I) .GE. COMP) COMP = TEMP(I)
20 CONTINUE
TEN(10) = COMP
TEN(1) = COMP/10
DO 25 I= 1,8
IC=10-I
25 TEN(IC) = IC*TEN(1)
DO 70 I= 1,5
NS1=0
DO 60 J= 1,9
IF(NS1 .NE. 0) GO TO 60
II = J+l
IF(TEMP(I) .GE. TEN(J) .AND. TEMP(I) .LE. TEN(II)) GOT TO 45
GO TO 60
45 IP = 10 -J
IF (ARY(IP) .EQ. TEMP(I)) GO TO 60
IF (ARY(IP) .GT. 0.) GO TO 50
ARY(IP) = TEMP(I)
NS1=1
GO TO 60
50 L = IP
TEMP1= TEMP(I)
DO 55 M = 1,10
IF (L .GE. 10) GO TO 55
IF TEMPII .EQ. ARY(L)) GO TO 55
IF TEMPII .GT. ARY(L)) GO TO 53
GO TO 54
53 IP = L
TEMP2 = ARY(L)
ARY(L) = TEMPI
TEMP1 = TEMP2
54 L = L+1
55 CONTINUE