The Effectiveness of Multi-Criteria Intelligence Matrices In Intelligence Analysis
Transcript of The Effectiveness of Multi-Criteria Intelligence Matrices In Intelligence Analysis
THE EFFECTIVENESS OF MULTI-CRITERIA INTELLIGENCE MATRICES IN INTELLIGENCE ANALYSIS
LINDSEY N. JAKUBCHAK
A Thesis
Submitted to the Faculty of Mercyhurst College
In Partial Fulfillment of the Requirements for
The Degree of
MASTER OF SCIENCEIN
APPLIED INTELLIGENCE
DEPARTMENT OF INTELLIGENCE STUDIESMERCYHURST COLLEGE
ERIE, PENNSYLVANIAMAY 2009
i
ii
DEPARTMENT OF INTELLIGENCE STUDIESMERCYHURST COLLEGE
ERIE, PENNSYLVANIA
THE EFFECTIVENESS OF MULTI-CRITERIA INTELLIGENCE MATRICES IN INTELLIGENCE ANALYSIS
A ThesisSubmitted to the Faculty of Mercyhurst CollegeIn Partial Fulfillment of the Requirements for
The Degree of
MASTER OF SCIENCEIN
APPLIED INTELLIGENCE
Submitted By:
LINDSEY N. JAKUBCHAK
Certificate of Approval:
_________________________________Kristan J. WheatonAssistant ProfessorDepartment of Intelligence Studies
_________________________________James BreckenridgeChairman/Assistant ProfessorDepartment of Intelligence Studies
_________________________________Phillip J. BelfioreVice PresidentOffice of Academic Affairs
MAY 2009
iii
iv
Copyright © 2009 by Lindsey N. JakubchakAll rights reserved.
v
“The work of managers, of scientists, or engineers, of lawyers-the work that steers the course of society and its economic and government organizations-is largely work of making decisions and solving problems. It is work of choosing issues that require
attention, setting goals, finding or designing suitable courses of action, and evaluating and choosing among alternative actions.”
-Nobel Laureate Herbert Simon
“Life is the sum of all your choices. History equals the accumulated choices of all mankind.”
-Albert Camus
vi
DEDICATION
This work is dedicated to:
-My parents for their continued support and confidence in me and everything I have done throughout my life.
-My friends who have greatly impacted the person that I am today.
-My classmates who are constantly pushing me to achieve new levels of academic and professional excellence.
vii
ACKNOWLEDGEMENTS
I would like to acknowledge my thesis advisor and primary reader, Kristan Wheaton, for
his guidance throughout the thesis process and throughout my graduate studies at
Mercyhurst College.
I would like to thank Professor Hemangini Deshmukh for her assistance with the
statistical analysis on this work.
I would also like to thank Justin Hiskey for his assistance with editing my thesis, and for
his overall support throughout my graduate studies
Finally, I would like to acknowledge Joshua Peterson for making his work available to
me, and for his advice throughout the process.
viii
ABSTRACT OF THE THESIS
The Effectiveness of Multi-Criteria Intelligence Matrices in Intelligence Analysis
By
Lindsey N. Jakubchak
Master of Science in Applied Intelligence
Mercyhurst College, 2009
Professor Kristan J. Wheaton, Chair
While there is a substantial body of literature related to the use and effectiveness
of Multi-Criteria Decision Making (MCDM) in the general sense, there is none regarding
the use of this method in intelligence analysis. This study discusses relevant literature on
the conventional form of MCDM and addresses the differences necessary to convert
MCDM to an intelligence methodology (tentatively titled Multi-Criteria Intelligence
Matrices-MCIM). Additionally, a controlled experiment was conducted to test the
hypothesis that MCIM is a valuable method to use in intelligence analysis. Findings
suggest that MCIM is likely a valuable method to use when conducting analysis, though
more research is recommended to confirm these results.
ix
TABLE OF CONTENTS
Page
COPYRIGHT PAGE …………………………………………………………... iii
QUOTES………………………………………………………………………... iv
DEDICATION…………………………………………………………………. v
ACKNOWLEDGMENTS……………………………………………………… vi
ABSTRACT……………………………………………………………………. vii
TABLE OF CONTENTS………………………………………………………. viii
LIST OF FIGURES…………………………………………………………….. xi
CHAPTER
1 INTRODUCTION……………………………………………… 1
2 LITERATURE REVIEW……………………………………….
Intuition vs. Structured Methodology……..……………. Benefits of Structured Methodology in Intelligence…….
Testing Structured Methodology in Intelligence…….…..MCDM………………………...…………………………The MCDM Process ……………………….……….…...
Validity of MCDM……………………………..…….…..Converting MCDM to MCIM……………………………Hypotheses ………………………………………………
4
468912182425
3 METHODOLOGY………………………………………………
Research Design………………………………………..Selection of Subjects…………………………………...
Recruitment of Subjects………………………………..Process…………………………………………………
Control Group…………………………………………. Experiment Group…………………………….............. Data Analysis Procedures……………………………..
27
27282830303133
x
4 RESULTS………………………………………………………..
Pre-Test Findings……..……………………………….....Post-Test Findings……………………………….……....Product…………………………………………...………Process……………………………………………………
Timeliness………………………………………………... Analytic Confidence…………………………………….. Objectivity and the Incorporation of Alternative Analysis………….. Logical Argumentation…………………………………. Bullet Point Analysis……………………………………. Potential of Future Use…………………………………..
36
363840424345
46464748
5 CONCLUSIONS………………………………………………...
Pre-Test and Post-Test Conclusions……………………... Process and Product Conclusions……………………….. A Step Forward…………………………………………..
50
505253
BIBLIOGRAPHY………………………………………………………………. 56
APPENDICES………………………………………………………………….. 59
APPENDIX 1: Analytic Methods Experiment Sign-Up Form….. 60
APPENDIX 2: Institutional Review Board Proposal Form…….. 61
APPENDIX 3: Participation Consent Form. …………………… 66
APPENDIX 4: Experiment Section 1 Handout (Control Group).. 67
APPENDIX 5: Analytic Methods Experiment Links…………… 69
APPENDIX 6: Pre-Test Questionnaire (Control and Experiment Groups) ………………………………………… 70
APPENDIX 7: Post-Test Questionnaire (Control Group)…………………….……………………. 72
APPENDIX 8: MCDM Participation Debriefing……………….
74
APPENDIX 9: MCDM Lecture Notes………………….……… 75
APPENDIX 10: Experiment Section 2 Handout (Experimental
xi
Group)………………………………………… 77
APPENDIX 11: Post-Test Questionnaire (Experimental Group)…………………………………………
APPENDIX 12: Statistical Data………………………………..
LIST OF FIGURES
80
82
xii
Page
Figure 1.1 General Example Matrix Used in MCIM 13
Figure 1.2 Step Two in the MCIM Process 15
Figure 1.3 Steps Three and Four in the MCIM Process 16
Figure 1.4 Steps Five and Six in the MCIM Process 17
Figure 1.5 Completed Example Matrix 18
Figure 3.1 Subjects by Class Year 29
Figure 3.2 Demographic Breakdown by Experiment Group 33
Figure 4.1 Control Group Results 41
Figure 4.2 Experiment Group Results 41
Figure 4.3 Length of Analysis Completion Time 44
Figure 4.4 Levels of Analytic Confidence 45
Figure 4.5 Bullet Point Analysis 48
xiii
1
CHAPTER 1:
INTRODUCTION
We all make decisions on a daily basis. In fact, we often make so many decisions
that we may fail to recognize we are even doing so; it is an act that has become second
nature to us. Some decisions, such as what we choose to eat for breakfast in the morning,
or which road we use to travel to work, are typically made with our “gut feeling” or
intuition. While these decisions may require the balancing of multiple factors (how
hungry we are or how much time we have), they require little forethought or analysis.
These decisions are not life altering, and they do not generally have a long-term affect on
us. On the other hand, decisions such as what kind of car we buy, or what career path we
embark on, are decisions that are affected by multiple criteria and may alter our life in
some way. In the field of intelligence, decisions made by others can affect matters of
national security, law enforcement and business operations. These decisions are directly
related to the safety and survival of both nations and organizations; therefore, they
mandate appropriate analysis.
Since the ability to make the “right” decision or predict an adversary’s most likely
course of action (COA) can have heavily influence the society in which we live, it is
imperative that an analyst makes sure that he or she fully grasps the relevant information
to provide effective analysis to a decision maker. One method of ensuring appropriate
evaluation of significant criteria, as well as the assessment of suitable alternatives, is
through the use of a structured methodology, such as MCDM.
2
MCDM is a generic term used to encompass a broad range of analytical
methodologies that use matrices as the basis for their conclusions. More specifically,
MCDM is an internally focused decision support method that assesses possible COAs
based on the evaluation of multiple criteria, goals, and objectives of conflicting nature. It
provides substance to a decision maker, as it allows for selecting an option based on the
appropriateness of those alternatives weighted against one another.
Given recent intelligence failures, substantial emphasis has been placed on
improving intelligence analysis and capabilities. In the Vision 2015 document, the
Director of National Intelligence specifically addresses the need to improve the
Intelligence Community (IC) and “create decision advantage.”1
One way to create decision advantage, as well as improve competitive analysis,
and counterintelligence capabilities, is by gaining insights into the thoughts of one’s
competitor or adversary through the use of Multi-Criteria Intelligence Matrices (MCIM).2
MCIM is a twist to the conventional form of MCDM, as it focuses not on an
organization’s own COA but rather on the likely COA of others outside the organization.
This method, once validated in an intelligence environment, might promote a more
efficient way of structuring intelligence data and may increase an analyst’s ability to
learn what is happening with his or her adversary or the environment in which they are
working in. Additionally, it may ensure that an analyst examines all relevant COAs
allowing for more thorough analysis of a situation. Finally, it may help an analyst escape
analytic pitfalls, and provide more credibility to an analyst’s work by providing a logical,
transparent process to a decision maker in the form of an easy to understand matrix.1 United States Government, Director of National Intelligence, Vision 2015, Available at
http://www.dni.gov/Vision_2015.pdf; Internet (Accessed 20 March 2009).2 MCIM is used to make the distinction between externally focused matrix analysis and MCDM,
or internally focused matrix.
3
While there is a substantial body of literature related to the use and effectiveness
of MCDM in the general sense, there is none regarding the use of this method in
intelligence analysis. The purpose of this experimental research is to determine the
effectiveness of using MCIM as an analytical methodology in the field of intelligence.
Currently no research has been conducted on the topic, and this study hopes to take the
first steps of many in increasing an analyst’s ability to thoroughly examine all possible
COAs in a timely manner as well as accurately forecast what COA one’s adversary is
likely to choose.
4
CHAPTER 2:
LITERATURE REVIEW
Some individuals will always argue that nothing can compete with the use of an
analyst’s “gut feeling” or intuition to make decisions, or that on occasion, there just isn’t
time to apply a structured methodology to an intelligence problem. After all, most
analysts are working under a deadline and cannot afford to incorporate time-consuming
methodologies that are difficult to use. However, in failing to implement a structured
method of problem solving, an analyst may be sacrificing quality analysis.
Intuition vs. Structured Methodology
In Michael R. LeGault’s book, Think: Why Crucial Decision Can’t Be Made in
the Blink of an Eye, he suggests that individuals should use their knowledge and abilities
to make informed decisions, rather than base decisions merely on intuition or impulse.
He further stresses the importance of critical or complex thinking by stating, “It can be
used simply to dig into things to enhance one’s awareness and discernment.”3 This quote
can be directly related to the analytic process in intelligence. When an analyst draws
conclusions based on their own experiences or intuition, he or she may fail to account for
conditions or factors in their adversary’s environment that may lead to a different
conclusion. By systematically reviewing all pertinent information, through the use of a
structured analytic method, an analyst may increase his or her chance of creating accurate
analytic judgments.
3 LeGault, Michael R, Think: Why Crucial Decision Can’t Be Made in the Blink of an Eye, New York, Threshold Editors, 2006, Page 305.
5
Ernest Forman, Professor of Management Science at George Washington
University, also highlights the benefit of structured thinking over intuitive thinking. He
states, “Although the vast majority of everyday decisions made intuitively are adequate,
intuition alone is not sufficient for making complex, crucial decisions.”4 He argues that
“organizations that use modern decision support methods can gain and maintain a
competitive edge in leading and managing global business relationships that are
influenced by fast changing technologies and complicated by complex interrelationship
between businesses and governments.”5 Forman’s position highlights the need for
teaching and implementing competitive decision making techniques, or structured
analytic methods.
Like Forman, an additional argument by LeGault demonstrates the importance of
teaching structured methodology early on in an analyst’s career. He states, “It [critical
thinking] is a skill that every individual is born with; however, it requires development
upon laying a foundation.”6 Therefore, if an analyst learns how to effectively use
analytical methodologies in the early stages of his or her career, they may expand and
develop the critical thinking skills necessary to make better decisions and more accurate
forecasts throughout their career. It is also important to recognize here that intuitive
decision making is not always available or beneficial to entry-level intelligence
professionals. These individuals often lack experience in their field, thus suggesting the
need for a more organized approach to deal with complex problems, in order to produce
better analysis.
Benefits of Structured Methodology in Intelligence4 Forman, Ernest, Decision by Objectives (How to Convince Others That You Are Right, Available
at http://mdm.gwu.edu/forman/DBO.pdf; Internet.5 Ibid.6 LeGault, Page 32.
6
Numerous experts in the field of intelligence have discussed the potential benefits
of using structured methodology in analysis. According to Stephen Marrin, “the primary
value of analytic techniques or structured methods is that they provide a way to account
for the analytic judgment; an analytic ‘audit trail’ as it were.”7 Thus, thru the
implementation of a structured methodology, the analyst’s thought process becomes more
transparent to others. Marrin also states, “With an analytic audit trail, analysts and their
colleagues can discover the sources of analytic mistakes when they occur and evaluate
new methods or new applications of old methods. In this way, structured methods make
it possible to advance the analytic tradecraft.”8 In other words, if the desired outcome
was not achieved, an analyst or decision maker can review what went wrong, or learn
how to approach and evaluate problems better in the future.
According to Richards Heuer, the use of structured methodology in intelligence
analysis may also help to prevent analytic pitfalls, such as the satisficing strategy or
groupthink.9 One structured methodology in particular, the Analysis of Competing
Hypothesis (ACH), requires analysts to weigh alternatives, in order to reach some type of
conclusion. Additionally, ACH provides a structured outline as to how an analyst
reached a specific conclusion, thus providing substance to a decision maker regarding his
or her estimates.
In regards to ACH, Heuer that stated the following benefits the methodology:
7 Marrin, Stephen, “Intelligence Analysis: Structured Methods or Intuition,” American Intelligence Journal, Page 7, Summer 2007.
8 Ibid, Page 7.9 Heuer, Richards, Psychology of Intelligence Analysis [book on-line] (Langley, Virginia: Central
Intelligence Agency Center for the Study of Intelligence, 1999), Available at https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf; Internet, Page 109.
7
This procedure [ACH] leads you through a rational, systematic process that avoids some common analytical pitfalls. It increases the odds of getting the right answer, and it leaves an audit showing the evidence used in your analysis and how this evidence was interpreted. If others disagree with your judgment, the matrix can be used to highlight the precise areas of disagreement. Subsequent discussion can then focus productively on the ultimate source of the differences.10
Thus, structured methods such as ACH provide analysts with a more efficient way to
approach and organize complex problems
Structured methodology may also help an analyst record all relevant pieces of
information. Sometimes, ideas and thoughts that analysts do not write down may get lost
or forgotten about. In fact, George A. Miller stated that “the number of things most
people can keep in working memory at one time is seven, plus or minus two.” 11 Thus, by
implementing a structured method, an analyst would have an organized way to keep track
of information so that it is not lost or forgotten.
Structured methodology in intelligence may also facilitate an analyst’s
understanding of how pieces of a problem are related. Heuer stated the following
regarding coping with complexity in analysis: “put all the parts down on paper or a
computer screen in some organized manner such as a list, matrix, map, or trees so that we
and others cans see how they interrelate as we work with them.”12 Thus, Heuer is
10 Heuer, Richards J., Psychology of Intelligence Analysis [book on-line] (Langley, Virginia: Central Intelligence Agency for the Study of Intelligence, 1999), Available at https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf; Internet, Page 109.
11 Miller, George. A., “The Magic Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information,” The Psychological Review, Vol. 63, No. 2 (March 1996). As cited by Heuer, Richards Jr. in “Taxonomy of Structured Analytic Techniques,” Paper prepared for the International Studies Association 2008 Annual Convention, March 26-29, 2008, San Francisco, CA.
12 Heuer, Richards, Jr., “Taxonomy of Structured Analytic Techniques,” Paper prepared for the International Studies Association 2008 Annual Convention, March 26-29, 2008, San Francisco, CA, Available at http://www.allacademic.com//meta/p_mla_apa_research_citation/2/5/4/1/2/pages254125/p254125-1.php; Internet; (Accessed 20 November 2008).
8
affirming the idea that structured methodology may provide some organized way for
analyst to visualize and evaluate data and see how pieces of criteria relate to each other.
Testing Structured Methodology in Intelligence
An experiment conducted in 2000 by MSgt Robert D. Folker, Jr. of the United
States Air Force (USAF) specifically addressed the use of structured methodology in
intelligence. The base of Folker’s research revolved around an experiment designed to
test the effectiveness of structured methodology in providing “correct” intelligence
analysis, by comparing it to intuitive analysis.13
Folker's subjects were various students at the Joint Military Intelligence College
(JMIC), which he divided into two groups: a control group and an experimental group.
His experimental group was trained on a structured methodology and asked to
incorporate that methodology in their analysis, while the control group was not asked to
use a specific methodology. Folker gave both groups the same two hypothetical
intelligence scenarios based off of facts from a real world problem. By forming the basis
of his experiment around an event that actually took place, Folker was able to assess an
analyst’s ability to pick the “correct” COA. Upon review of the results, Folker’s found
that the group of analysts, who used structured methodology in their analysis, provided
significantly better analysis than the group who used no formal structured methodology.14
He also concluded that these results were not linked to “rank, experience, education or
13 Folker, MSgt Robert D, Jr., “Intelligence Analysis in Theater Joint Intelligence Centers: An Experiment in Applying Structured Methods,” Joint Military Intelligence College, Washington DC, NDIC Press 2000. Available at http://www.fas.org/irp/eprint/folker.pdf; Internet, Page 15
14 Ibid, Page 32
9
branch of service.”15 Folker further stated, in order to be effective, intelligence analysts
must be properly trained on the specific type of methodology.16
Folker’s research was important as it reinforced the idea that the implementation
of structured methodology can assist analysts in breaking down problems and producing
better analysis.
MCDM
In the Psychology of Intelligence Analysis, Richards Heuer, also distinguishes
between merely “sitting down and thinking about a problem and really analyzing a
problem.”17 Like many others, he suggests that in order to fully analyze a complex
problem, an individual must break it down into its’ smaller components, evaluate each
portion, and then piece the subcomponents back together to make a decision.18 One
method of doing just that is through the use of decision support systems such as MCDM.
As noted earlier, MCDM is a generic term used to encompass a broad range of
analytical methodologies that use matrices as the basis for their conclusions. More
specifically, MCDM is an internally focused decision support method that assesses COAs
based on the evaluation of multiple criteria, goals, and objectives of conflicting nature.
Prior to demonstrating the validity of MCDM, as well as judging the effectiveness of
MCIM, it is necessary to begin by providing a brief overview of the traditional process.
As noted earlier, the conventional process of MCDM has numerous variations.
15 Folker, Page 3216 Folker, Page 3217 Heuer, Richards, Psychology of Intelligence Analysis, [book on-line] (Langley, Virginia: Central
Intelligence Agency Center for the Study of Intelligence, 1999), Available from https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf: Internet, Page 94.
18 Ibid, Page 94.
10
Dr. Thomas Saaty developed the Analytic Hierarchy Process (AHP), one variation
of MCDM. AHP allows decision makers to structure decisions hierarchically to
determine which one most suits their needs. Typically, the overall goal or decision to be
made, is stated at the top of the hierarchy, criteria used evaluate the alternatives are in the
middle, and the various COA are listed at the bottom. In regards to the benefits of his
methodology Saaty states:
The AHP is about breaking a problem down and then aggregating the solutions of all the sub problems into a conclusion. It facilitates decision making by organizing perceptions, feelings, judgments, and memories into a framework that exhibits the forces that influence a decision. In the simple and most common case, the forces are arranged from the more general and less controllable to the more specific and controllable.19
Another variation of MCDM is the WSM. According to research conducted by
Evangelos Triantaphyllou, the WSM is “probably the most commonly used approach,
especially in single dimensional problems.”20 Additionally, Triantaphyllou states that this
method works best “in single dimension problems where all the units are the same, (e.g.
dollars, feet, seconds)” and “difficulty arises when it is applied to multi-dimensional
decision making problems.”21
The WPM is another popular MCDM variation that is similar to the WSM model.
In fact, the main factor that distinguishes the WPM from the WSM is that the WPM
19 Saaty, Thomas L, “How To Make A Decision: The Analytic Hierarchy Process,” Available at http://sigma.poligran.edu.co/politecnico/apoyo/Decisiones/curso/Interfaces.pdf; Internet (Accessed 24 March 2009).
20 Ibid, Page 6.21 Triantaphyllou, Evangelos and Stuart H. Mann, “An Examination of the Effectiveness of Multi-
Criteria Decision Making Methods: A Decision Making Paradox,” Available At http://www.csc.lsu.edu/trianta/index.html?http://www.csc.lsu.edu/trianta/Books/DecisionMaking1/Book1.htm; Internet (Accessed 20 March 2009).
11
applies multiplication to criterion, in order to compare alternatives, where as the WSM,
uses addition.22
The staff study is also a MCDM variant that is commonly studied and utilized in
the military community. As outlined in the United States Army Field Manual (FM) 101-
5, Staff Organizations and Operations, there are several types of military briefs including
information, decision, mission and staff.23 FM 101-5 states that the decision brief,
designed to arrive at the solution to a problem/decision, “requires that an analyst be
prepared to present his assumptions, facts, alternative solutions, reason for choosing the
recommended solution, and the coordination involved.” The staff study format, outlined
in Appendix D of FM 101-5,24 provides the means to do this by outlining multiple COAs,
reviewing pertinent criteria, and evaluating the most appropriate solution.
22 Triantaphyllou, Evangelos and Panos M. Parlos (ed.), Multi-Criteria Decision Making Methods: A Comparative Study, The Netherlands: Kluwer Academic Publishers, 2000, Page 8.
23 The United States Army. Field Manual 101-5. Staff Organization and Operations. 31 May 1997. Page 115. Available at http://www.fs.fed.us/fire/doctrine/genesis_and_evolution/source_materials/FM-101-5_staff_organization_and_operations.pdf; Internet.
24 Ibid, Page 105.
12
The MCDM Process
Although MCDM is an umbrella term used for a collection of methodology types,
they all involve the idea of decision alternatives, or COAs, determined by numerous
criteria weighted against each other. In MCDM, emphasis is on placing value, or
judgment of an item’s worth and desirability, on pieces of criterion, a standard on which
a judgment or decision may be based.25 By providing a breakdown of the conventional
MCDM process, I hope to demonstrate that this methodology is relatively easy to
understand, and when used properly, it is an effective way to approach complex
problems.
For this research, I have chosen to use a MCDM model derived from the Army’s
staff study. This model seemed like an appropriate choice given its simplicity and ability
to accommodate a large amount of unstructured data, which is common to intelligence
analysis.
In general, the basic parts of the MCDM process can be broken down into the
following eight steps:
1. Requirement/Question and Collection2. Establish Possible COAs3. Establish Screening Criteria4. Screen COAs5. Establish Evaluation Criteria6. Weight Evaluation Criteria7. Evaluate COAs8. Make Recommendations
25 Webster’s Dictionary Online, Available at http://www.merriam-webster.com/dictionary/criteria; Internet.
13
Requirement/Question and Collection
The first step in MCDM, establishing a clearly defined question, is essential. An
analyst simply cannot have a good analysis without a clearly defined question. In this
step, it is equally important that the analyst or decision maker form the question in a way,
which will elicit possible COAs. For the purpose of working through the conventional
process of MCDM, I have chosen to explain each step by addressing the question of
which vehicle to buy.
MCDM normally uses a matrix to display the results of the analysis. The general
matrix used in MCDM normally looks like the one below, where COA represents the
various options realistically open to the decision maker and criteria represents the various
relevant ways in which a decision maker will judge each COA.
Figure 1.1 Example of a General Matrix Used in MCDM
Question: Which kindof vehicle should I buy?
Criteria 1:
Criteria 2:
Criteria 3:
Criteria 4:
Total/ (RankOrder):
COA 1:
COA 2:
COA 3:
COA 4:
COA 5:
COA 6:
14
Establish Possible COA
The second step in the MCDM process is establishing relevant possible COAs.
According to numerous experts, there are two categories for further classifying MCDM
problems based on the possible COAs: Multi-Objective Decision Making (MODM) and
Multi-Attribute Decision Making (MADM).26 The primary difference between these two
categories is that with MODM problems, the decision alternatives are endless; whereas,
MADM problems focus on distinct decision spaces, or predetermined COAs.27 In the
case of deciding what car to buy, the example question falls in line with the idea of
MODM, as there are endless possibilities of cars for an individual to choose from, and an
individual is not required to choose from pre-determined COAs. In this step, it is
important for an analyst to identify as many COAs as possible, since there is no “correct”
number of alternatives. The range of possible COAs merely derives from the extent of
the problem and the creativity of the analysts in solving the problem.
The following example matrix outlines six types of vehicles that an individual
may purchase. It is important to note that this is not an all inclusive list, and is merely an
abridged list for the purpose of demonstrating the use of matrices.
26 Triantaphyllou, Evangelos and Panos M. Parlos (ed.), Multi-Criteria Decision Making Methods: A Comparative Study, The Netherlands: Kluwer Academic Publishers, 2000, Page 1.
27 Ibid, Page 1.
15
Figure 1.2 Step Two in the MCDM Process
Question: Which kindof vehicle should I buy?
Criteria 1:
Criteria 2:
Criteria 3:
Criteria 4:
Total/ (RankOrder):
COA 1:Family compact carCOA 2:Mini VanCOA 3:Sports carCOA 4:Station wagonCOA 5: SUVCOA 6:Motorcycle
Establish Screening Criteria and Screen Criteria
After completion of the second step in the MCDM process, an individual might be
overwhelmed at the plethora of COAs available; therefore, he or she must begin to
eliminate certain COAs by establishing screening criteria. It is necessary to distinguish
this step separate from the previous step, which involved establishing possible COAs.
On occasion, individuals subconsciously rule out possible COAs, without even
recognizing they are going through a screening process. For example, when choosing
which car to buy, a student may not even consider options that may be out of range for
either financial reasons (too costly) or due to appearance, (it is not a dream car),
thus,ruling out possible alternatives before having the opportunity to compare it against
other COAs.
In the example case, screening criteria for our aforementioned problem may be
“must be vehicle two wheeled vehicle.” By establishing “must have/be” guidelines for
COAs an individual can quickly eliminate unlikely possible COAs. Based on the
16
previous evaluation criteria, the possible COA, “motorcycle” may be eliminated as it is
not a four wheel vehicle. The following matrix shows the elimination of one possible
COA (a motorcycle), after implementing the screening criteria.
Figure 1.3 Steps Three and Four in the MCDM Process
Question: Which kindof vehicle should I buy?
Criteria 1:
Criteria 2:
Criteria 3:
Criteria 4:
Total/ (RankOrder):
COA 1:Family Compact CarCOA 2:Mini VanCOA 3:Sports CarCOA 4:Station WagonCOA 5: SUVCOA 6:Motorcycle
X X X X X
Establish Evaluation Criteria and Weight Evaluation Criteria
The next steps in the process, establishing evaluation criteria and weighting the
evaluation criteria, allows the analyst to provide rank to the remaining COAs. At this
point, the analyst uses evaluation criteria to establish that “all things being equal, this is
the best COA for me." The analyst should develop a system of rank that stresses the
significance of one criterion over another. The analyst can provide rank as either a
tangible means, things that he or she can measure, or by intangible means, things that he
or she cannot easily measure.
Evaluation criteria for the stated problem may be specific, “must cost less than
$25,000,” or it may be generalized, “with all other things being equal, the most
17
inexpensive car is the best.” This step involves evaluating what an analyst or decision
maker really cares about, or what is most important in achieving the desired outcome.
Additionally, this step allows an analyst to provide a system to weight the most important
criteria to them. For example, if the individual was a student of limited financial means,
price may be the most important criteria when looking to buy a vehicle. Therefore, the
price category will be multiplied by two, in order to demonstrate its significance. The
following matrix outlines the criteria, while applying weight and rank order:
Figure 1.4 Steps Five and Six in the MCDM Process
Question: Which kindof vehicle should I buy?
Criteria 1:PriceCheaper isBetter(x2)
Criteria 2:SizeRoom forfamily
Criteria 3:SafetyForFamily
Criteria 4:AppealAttractiveLook
Total/ (RankOrder):
COA 1:Family Compact Car
1 (2) 4 1 2 9
COA 2:Mini Van
2 (4) 1 2 4 11
COA 3:Sports Car
5 (10) 5 5 1 21
COA 4:Station Wagon
3 (6) 3 3 5 17
COA 5: SUV
4 (8) 2 4 3 17
Evaluate COA and Make Recommendations
After establishing evaluation criteria, it is necessary to provide a system of rank
order to the possible COAs by establishing a scale. When addressing our example
question, this task could be as simple as ranking the cars most likely to purchase, to least
18
likely to purchase, by utilizing a scale from 1 to 3, 1 to 5, or 1 to 10. The following
completed matrix highlights the rank order of the possible COAs:
Figure 1.5 Completed Example Matrix
Question: Which kindof vehicle should I buy?
Criteria 1:PriceCheaper isBetter(x2)
Criteria 2:SizeRoom forfamily
Criteria 3:SafetyForFamily
Criteria 4:AppealAttractiveLook
Total/ (RankOrder):
COA 1:Family Compact Car
1 (2) 4 1 2 9 (1)
COA 2:Mini Van
2 (4) 1 2 4 11 (2)
COA 3:Sports Car
5 (10) 5 5 1 21 (5)
COA 4:Station Wagon
3 (6) 3 3 5 17 (3)
COA 5: SUV
4 (8) 2 4 3 17 (3)
After ranking the possible alternatives, the individual now has a structured format,
which provides substance to their decision. The family compact car appears to be the
most idea vehicle, followed by the minivan, station wagon and SUV, and sports car.
Upon completion of the first MCDM analysis, one COA that may be chosen is a
sports car. It is important to note, that upon determining the first ideal COA, a second
matrix analysis may be completed in order to determine the more specific type of sports
car.
Validity of MCDM
19
In his book, Multi-Criteria Decision Making Models: A Comparative Study
(2000), Evangelos Triantaphyllou explores some of the variations on MCDM to
determine what the most “ideal” variation is to use. Since it was not possible to include
every possible MCDM variation, Triantaphyllou focused on the more popular methods,
such as AHP, WSM, and the WPM, noted above. Although unrealistic to determine the
most “ideal” method to use in a given situation, Triantaphyllou’s extensive research of
the variations of MCDM provides strong evidence that there is a plethora of methods
available which allow an efficient and organized approach to effective decision making.
Triantaphyllou also points out the irony in his research commenting, “Deciding on which
one is the best method can be viewed as an MCDM problem itself whose solution
requires the use of the best MCDM.”28
Valerie Belton and Theodor Stewart also provide an excellent overview of the
various types of MCDM models and theories that have developed over the past 25 years,
in their book Multiple Criteria Decision Analysis: An Integrated Approach (2002). The
authors reinforce the idea that although MCDM cannot guarantee a “correct” answer, the
methodology should “facilitate decision makers’ learning about the understanding of the
problem faced, about their own, other parties’ and organizational priorities, values and
objectives and through exploring these in the context of the problem to guide them in
identifying a preferred COA.”29 The book concludes with the authors’ opinion regarding
the future of MCDM, specifically stating that integration of various perspectives and
28 Triantaphyllou, Evangelos, “Can We Always Determine the Right Alternatives in Business Problems,” 18 August 2002.
29 Belton, Valerie and Theodor J. Stewart, Multiple Criteria Decision Analysis: An Integrated Approach, The Netherlands: Kluwer Academic Publishers, 2002, Pages 2-3.
20
identifying common strengths and weaknesses, is essential to the methodology’s “growth
and success.”30
Aside from the research surrounding MCDM as outlined in the aforementioned
academic books, MCDM has multiple applications in the real world, contributing to such
fields as finance and economics, environmental management and marketing.31 In his
article “Decision making with the analytic hierarchy process,” Saaty justifies the
application of AHP by citing examples where the methodology has been used in the real
world.
The following examples are just a few of the ones outlined by Saaty, and are
taken directly from his publication:
British Airways used the AHP in 1998 to choose the entertainment system vendor for its entire fleet of airplanes.
In 2001, the methodology was used to determine the best relocation site for the earthquake devastated Turkish city Adapzari.
Xerox Corporation has used AHP to allocate close to a billion dollars to its research projects.
In sports, it was used in 1995 to predict which football team would to go to the Superbowl and win. The AHP was also applied in baseball to analyze which players should be retained on a team.
In 1999, the Ford Motor Company used the AHP to establish priorities for criteria that improve customer satisfaction. Ford gave Expert Choice Inc, an Award for Excellence for helping them achieve greater success with its clients.32
The application of MCDM problems has also surfaced in financial planning
journals. In the March 2008 Journal of Financial Planning, William Z. Suplee and
Steven R. Dzubow discuss the importance of using structured methodology in helping
30 Belton and Stewart, Pages 333-343.31 Zopounidis, Constantin and Doumpos, Michael, “Multiple-Criteria Decision Making,” Thomson
Corporation, 2006. 32 Satty, Thomas , “Decision making with the analytic hierarchy process,” Int. J. Services
Sciences, Vol. 1, No. 1, 2008, Available at http://inderscience.metapress.com/media/pgwf2qtuyg3xnvnhxnby/contributions/0/2/t/6/02t637305v6g65n8.pdf; Internet (Accessed 26 March 2009).
21
their clients make critical financial decisions. In their article, “Using Multiple-Criteria
Decision Analysis to Simply the Financial Planning Process,” they demonstrate the value
of using matrix analysis by a parent and child deciding on the best college for the child to
attend through the process of evaluating a plethora of criteria including tuition costs,
graduation rate, campus life, student/teacher ratio and post-graduation placement.33
In addition to financial journals, the relevance of MCDM has also appeared in
environmental management journals. In their article, the “Application of Multicriteria
Decision Analysis in Environmental Decision Making,” in The Integrated Environmental
Assessment and Management Journal, Gregory A. Kiker, Todd S. Bridges, Arun
Varghese, Thomas P. Seager and Igor Linkov discuss the application of MCDM in
various environmental projects. The authors highlight numerous environmental case
studies, by outlining the specific MCDM method used, the decision context (a range of
topics in the areas of environmental management, stakeholder involvement and
contaminated sites) and the funding agency (a range of universities, as well as US and
international government agencies).34 This article is mentioned as it demonstrates the
flexibility and use of MCDM in a multiplicity of environmental topics conducted by an
array of organizations.
A surplus of research and studies has also been conducted as to the application
and relevance of MCDM methods in engineering. In the fields of civil and
environmental engineering, the Elimination and Choice Translating Reality (ELECTRE)
33 Dzubow, PhD and William Z. Suplee IV, CFA, CFP, ChFC, CAS, “Using Multiple-Criteria Decision Analysis to Simplify the Financial Planning Process,” Journal of Financial Planning. March 2008, Available at http://www.library.idsc.gov.eg/GUI/Globals/Upload/BULLETIN_ATTACHMENT/92/e-files/manegement%20and%20economics/using%20multiple.pdf; Internet ( Accessed 5 December 2008).
34 Kiker, Gregory A., Todd S. Bridges, Arun Varghese, Thomas P. Seager and Igor Linkov. “Application of Multicriteria Decision Analysis in Environmental Decision Making.” Integrated Environmental Assessment and Management. Volume 1, Number 2, pages. 95-108. 2005, Available athttp://www.allenpress.com/pdf/ieam-01-02_95_108.pdf; Internet. (Accessed 20 November 2008).
22
II and III, are cited as two MCDM variations that are widely used.35 While conducting
his research in regards to MCDM in engineering, Xiaoting Wang mentions several past
case studies in which MCDM methods were applied. Once again, the variety of topics
covered highlights the flexibility in the application of MCDM problems. Some of the
real life case studies utilized in Wang’s research included “choosing a solid waste
management system [Hokkanen, J., and P. Salminen (1997)], choosing an alternative fuel
system for land transportation [Poh, K.L., and B.W. Ang, (1999), and selection of an
alternative electricity power plant [Leyva-Lopes, J.C., and E. Fernandez-Gonzalez
(2003)].”36
In their article “Multi-Criteria Decision Making Process for Buildings,” J.D.
Balcomb and A. Curtner demonstrate MCDM’s value in the building design planning
process. The authors suggest that by evaluating criteria such as life cycle cost,
architectural quality, functionality, air quality, and maintainability, various building
designs can be evaluated to ensure maximum sustainability. The authors indicate that as
fundamental decisions are being made, the methodology has the ability “to assist design
teams in prioritizing their goals, setting performance targets and evaluating design
options to ensure that the most important issues affecting building sustainability are
considered.”37 Additionally, the paper highlights the various strengths of the
methodology including ease of use, its inexpensive nature, its ability to document how
35 Wang, Xiaoting, “Study of Ranking Irregularities When Evaluation Alternatives By Using Some ELECTRE Methods and a Proposed New MCDM Method Based on Regret and Rejoicing,” A Thesis Submitted to the Graduate Faculty of the Louisiana State University and Mechanical College. August 2007. Available online at http://etd.lsu.edu/docs/available/etd-07112007-012708/unrestricted/Wang_thesis.pdf; Internet (Accessed 24 March 2009), Page iv.
36 Ibid, Page 25.37 Balcomb, J.D. and A. Curtner, “Multi-Criteria Decision-Making Process for Building,” paper to
be presented at the American Institute of Aeronautics and Astronautics Conference, Las Vegas, Nevada, July 24-28. 2000. Available online at http://www.nrel.gov/docs/fy00osti/28533.pdf; Internet (Accessed 23 March 2009), Page 1.
23
the team arrived at conclusions, as well as its ability to provide visual representation of
the team’s thought process which can be interpreted by almost anyone.38
Research conducted by Stacy Gilchrist, Master of Science in Applied Intelligence
from Mercyhurst College, examined the use of the military staff study in intelligence by
transforming the conventional problem solving process into an intelligence-focused
methodology.39 For the purpose of his research, Gilchrist applied the methodology to the
following question regarding the Mara Salvatrucha gang (MS-13): “How will MS-13
respond or react to the current law enforcement efforts led by the Immigration and
Customs Enforcement (ICE) branch of the Department of Homeland Security (DHS),
specifically Operation Community Shield?”40
Gilchrist noted that first and foremost, the intelligence-focused staff study was
decision maker focused and provided a means to track the analyst’s thought process.41 A
second strength noted of the intelligence-focused staff study by Gilchrist is that the
process is structured and organized.42 A final strength that Gilchrist noted was that the
intelligence-focused staff study produced objective analysis, stating that cognitive biases
are likely to occur less frequently, due to the individual evaluation of each criterion in
regards the overall objective.43 Gilchrist also noted possible weaknesses including the
potential for the methodology to be both time-consuming and contain flawed results.44
38 Balcomb, J.D. and A. Curtner, Pages 2 and 8.39 Gilchrist, Stacy. “Staff Study,” (Research Paper for Advanced Analytic Techniques,
Mercyhurst College).40 Ibid, Page 3.41 Ibid, Page 1.42 Ibid, Page 343 Ibid, Page 1.44 Ibid, Page 1.
24
Gilchrist’s findings are important as they address the issues that many analysts
face when working with structured methodology. Additionally, Gilchrist’s research
provides valuable insight to the use of the intelligence-focused MCIM matrix.
Converting MCDM to MCIM
Since the conventional form of MCDM has proven to be valuable, why not
replicate the process, while making slight modifications, in order to create a valuable
structured methodology that may lead to the production of better intelligence analysis?
The only difference with using the methodology in intelligence is that the analyst would
apply the method to an external entity-to best predict the likely COA an enemy or
competitor will choose.
MCIM would likely have several benefits for intelligence analysts. Just as Heuer
states in regards to ACH, although the use of MCIM can never guarantee a “correct”
answer, it may increase the odds of choosing the COA one’s adversary or competitor is
most likely to pursue, based on a thorough breakdown of all relative pieces of
information.45 Additionally, the analyst can provide the decision maker with a supporting
matrix that outlines their thought process, which may give more credibility to their work.
The matrices can also highlight the areas of disagreement or conflict between two
individuals, by identifying the weight of importance that an analyst places on each
45 Heuer, Richards J., Psychology of Intelligence Analysis [book on-line] (Langley, Virginia: Central Intelligence Agency for the Study of Intelligence, 1999), Available at https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf; Internet (Accessed 11 November 2008), Page 109.
25
individual criterion.46 Finally, a decision maker or analyst can adjust the matrix to
accommodate new data and “what-if” scenarios.
Assuming the role of an external entity would also benefit intelligence analysis.
In fact, this is something the army has already been doing for years through the Red
Team methodology, a tactic frequently used to increase the odds of predicting how one’s
opponent is likely to react in a specific situation.47 Additionally, in his book, Strategy-
Specific Decision Making: A Guide for Executing Competitive Strategy, William Forgang
states that “external analysis has the ability to deter an organization from pursuing an
option they might have otherwise chosen to accept.”48 Thus, by using structured methods
with an external focus such as MCIM, an analyst might gain insight to factors outside of
a decision maker’s immediate surroundings, and possible deter them from costly or poor
decisions they would otherwise have chosen to pursue.
Hypotheses
The following experiment seeks to demonstrate that MCIM can prove itself as a
useful intelligence analysis methodology. Specifically, I hypothesized the following:
Hypothesis 1: The experimental group using MCIM to conduct analysis will
have a more accurate product than the students in the control group.
46 Heuer, Richards J., Psychology of Intelligence Analysis [book on-line] (Langley, Virginia: Central Intelligence Agency for the Study of Intelligence, 1999), Available at https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf; Internet, Page 109.
47 2008 U.S. Army Posture Statement. Information Papers: Red Team Education and Training, Available at http://www.army.mil/aps/08/information_papers/prepare/Red_Team_Education_and_Training.html; Internet (Accessed 26 September 2008).
48 Forgang, William G., Strategy-Specific Decision Making: A Guide for Executing Competitive Strategy. M.E. Sharpe, Inc., Armonk, New York, 2004.
26
Hypothesis 2: The analytical process for the experimental group using MCIM
will be more transparent than the control group.
Hypothesis 3: The experimental group using MCIM will complete their analysis
faster than the control group.
Hypothesis 4: The experimental group using MCIM will have a higher level of
analytic confidence in their product than the control group.
Hypothesis 5: The experimental group using MCIM will be more objective
during the analytic process and examine more alternative possibilities.
Hypothesis 6: The experimental group using MCIM will provide better logical
argumentation than the control group.
27
CHAPTER 3:
METHODOLOGY
In order to test the given hypotheses, I designed an experiment to assess whether
or not MCIM methodology improves intelligence analysis. When the experiment was
completed, I evaluated the analyst’s process and product, including the length of time it
took the student-analysts to complete analysis and their level of analytic confidence. This
methodology section will detail the research design of this experiment.
Research Design
The experiment I conducted broke the subjects into two groups. One of the
groups was an experimental group (comprised of 24 student-analysts), and one was a
control group (comprised of 21 students-analysts). I gave the control group a 10 minute
instruction/information period regarding the experiment tasking, and then asked them to
complete an analysis of a real life intelligence scenario. I taught the experimental group
the method of MCDM/MCIM over the course of approximately 35 minutes, gave them a
10 minute instruction/information period regarding the experiment tasking, and then
asked them to provide an analysis regarding the same real life intelligence scenario. The
only difference between the two groups was that the experimental group received
instruction regarding MCDM/MCIM and then was asked to complete the analysis using
that methodology. I gave the two groups access to the same resources and provided them
the same amount of time to complete their analysis.
28
Selection of Subjects
I used students at the Mercyhurst College Institute for Intelligence Studies
(MCIIS) as my research subjects. Mercyhurst College, located in Erie, Pennsylvania, is
the originator of a four-year program specifically designed to prepare students for entry-
level careers in business, law enforcement and national security intelligence analysis
Additionally, Mercyhurst College offers a master’s program, with a thesis requirement.
Since the students of Mercyhurst College are training to be efficient analysts, they could
arguable serve as a proxy for entry-level analysts in the intelligence community, who
would have been difficult, (if not impossible), to recruit due to security clearances and
access problems.
Recruitment of Subjects
In order to recruit participants for my experiment, I approached several professors
in the intelligence department asking for permission to recruit student-analysts during
their designated class time. Upon receiving their approval, and approximately one and a
half weeks prior to the start of my experiment, I gave a brief oral presentation at the
beginning of several classes that included information regarding the time, place and
subject matter of the experiment. The students were informed that the experiment would
be conducted during the second week of winter term. The presentation also indicated that
several professors in the department would be offering extra credit for their class if
students chose to participate in the experiment. Offering extra credit may have persuaded
several students to participate when they otherwise might not have. Upon agreeing to
participate, each student-analyst was handed a sign-up sheet (see Appendix 1) and asked
29
to choose the appropriate date and time that they were available to participate. The
students could choose from either a Wednesday or Thursday night time slot based on
their availability or preference. The time frame on both days was kept the same so that
the student-analysts were unable to determine which group was the control group and
which group was the experimental group. If students did not have a preference for
availability, they selected a third choice, in which they were randomly assigned a day to
complete their experiment.
The experiment was open to all students in the intelligence department, including
both undergraduate and graduate students (See Figure 3.1 for breakdown by of the
subjects by class year). A total of 45 students participated in experiment. Although the
experiment was open to all class years, no freshmen participated in the experiment.
Figure 3.1 Subjects by Class Year
30
Process
Prior to conducting my experiment, I needed approval from the Mercyhurst
College Institutional Review Board (IRB). The IRB requires any student anticipating
conducting an experiment involving human subjects to submit a proposal for review (See
Appendix 2). The purpose of the experiment proposal is to outline the experiment and
identify any possible risks, if any, which participants may come into contact with as a
result of participating in the experiment.
After receiving approval from the IRB, I also had to secure consent from
participants in the experiment. At the beginning of each experiment session, I outlined
what the experiment entailed. I then provided each student-analyst with a participation
consent form (See Appendix 3). The consent form asked for the student’s name, class
year, and name of their professor/professors, in order for me to pass along the appropriate
information to award them extra credit. Before starting each group’s section of the
experiment, I answered any questions that the subjects may have had regarding the
experiment. The process of each group was carried out the same way, only adjusting
those items necessary to test the effectiveness of the methodology.
Control Group
The first group of participants that I used to test my hypotheses on was the control
group. When the sessions started, I provided the group with a brief set of instructions and
provided them with a handout regarding a real life intelligence scenario. The handout
provided the subjects with room to write their analysis, as well as a section to identify
their level of analytic confidence (See Appendix 4). I provided the subjects with a
31
handout containing a list of websites to use as a starting point for their research (See
Appendix 5). I informed the students that these links were merely a starting point for
their analysis, and it was their choice if they wanted to use the links for information. I
asked the students to carefully read through the problem and gain an understanding of the
requirements. I then handled any questions they had regarding the process of the
experiment. After gaining an understanding of the expectations, and prior to completing
their tasking, I asked the subjects to complete a pre-test questionnaire and provide brief
demographic information, both of which I used to further compare the experimental and
control groups (See Appendix 6). I then gave the group two hours to complete their
analysis using whichever sources and method they preferred. Upon completing their
analysis, I asked the control group to complete a post-test questionnaire (See Appendix
7), and then provided them with a debriefing form (See Appendix 8).
Experimental Group
The second group of participants that I used to test my hypothesis was the
experimental group. At the beginning of the session, the experimental group received a
35 minute long presentation on MCDM/MCIM methodology, where the subjects worked
through examples of MCDM/MCIM problems. In addition, the group received a handout
on MCDM/MCIM that they were able to refer back to later on in the experiment (See
Appendix 9). I gave the experimental group the same 10 minute instruction period as the
control group, regarding the tasking. In addition, I provided them with a handout (See
Appendix 10) which asked for each subject to provide analysis for the same intelligence
related problem within a period of two hours. The only difference between the two
32
groups was that I asked the experimental group to complete their analysis using MCIM.
The handout provided room for the student-analysts to write their analysis and to indicate
their level of analytic confidence; however, it also instructed each subject to create an
MCIM matrix in Excel format. I provided the experimental group with an example of
what the matrix should resemble when their product was completed. Additionally, they
had the option of completing an outline, which asked for possible COAs, a listing of
screening criteria, the eliminated possible COAs and the evaluation criteria. After
receiving instruction, but prior to completing the requirement, I asked the experimental
group to complete the same pre-test questionnaire as the control group. Upon completion
of their analysis, I asked each member of the experimental group to fill out a post-test
questionnaire (See Appendix 11), which differed slightly from the control group, and
received the same debriefing form.
In order to obtain minimal outside influence with the results, I intended to keep
the control and experimental group as similar as possible as far as number of participants
and class year (See Figure 3.2 for a breakdown demographic breakdown of subjects by
group). Although the number of participants in each group was similar, (a total of 21
student-analysts participated in the control group and a total of 24 student-analysts
participated in the experimental group section), there was an unpreventable discrepancy
regarding the number of students in each class year. Specifically, there were a greater
number of first year graduate students within the experimental group. This was due to
uncontrollable scheduling conflicts regarding evening classes and meetings, as well as the
availability of students who wanted to participate in the experiment. In order to identify
if this discrepancy had any effect on the results, specific questions related to familiarity
33
and experience with analytic methods were addressed in both the pre-test and post-test
questionnaires.
Figure 3.2 Demographic Breakdown by Experiment Group
Data Analysis Procedures
Since the intent of the experiment was to test whether MCIM is a valuable
methodology to use in the field of intelligence, I evaluated several factors upon
completion of the experiment. Prior to the review of each analyst’s process and product,
it was necessary for me to draw some conclusions from both the pre-test and post-test
questionnaires in order to identify possible outside influence. A majority of the questions
required the students to rate their answer on a 5 level Likert scale. I recorded the number
of responses for each numeric category in a spreadsheet, found the percentage of subjects
34
who answered each number within each group, and figured out the average of both
groups. By comparing the averages, I was able to pinpoint any differences between the
two groups and test whether or not these differences were statistically significant at the
5% level.
In addition to these questionnaires, I asked each student-analyst to indicate their
level of analytic confidence. When determining his or her level of analytic confidence,
each student-analyst was asked to consider the following seven factors:49
1. Use of Structured Method (s) in Analysis2. Overall Source Reliability3. Source Corroboration/Agreement4. Level of Expertise on Subject/Topic & Experience5. Amount of Collaboration 6. Task Complexity7. Time Pressure
In order to measure analytic confidence, I asked the students to rate their level of
analytic confidence by marking their level of confidence on the modified scale shown
below.50
Low |------------------------------------------------------| High
The scale was 10 centimeters in length, and they placed a slash mark where they
felt their level of analytic confidence fell. In order to compare levels of analytic
confidence between groups, I measured each analyst’s slash mark to the nearest half
centimeter. I then determined the minimum and maximum level of analytic confidence in
the experimental group and control group, as well as the average level within each group.
49 Peterson, Joshua. “Appropriate Factors to Consider When Assessing Analytic Confidence in Intelligence Analysis.” (Master's Thesis, Mercyhurst College, March 2008)
50 Ibid
35
Upon completion of their analysis, I also asked each student-analyst to record the
amount of time it took them to complete their analysis (in minutes). In order to make
comparisons regarding length of time to complete their analysis, I found the average
amount of time, as well as the minimum and maximum amount of time within each
group. Finally, I reviewed each individual analysis, in order to make comparisons
regarding the thoroughness of their process, and the accuracy of their product.
36
CHAPTER 4:
RESULTS
Statistical analysis was conducted on the results in order to determine if the
findings were significant. If the results indicated a P-value < 0.05, then the results were
considered significant, meaning it is unlikely to have occurred by chance. If the results
indicated a P-value > 0.05, then the results were considered not significant, meaning it
was likely to have occurred by chance. This section will discuss the results of both the
pre-test and post-test questionnaires, as well as the results of the control group and
experimental group viewed both individually and compared as a whole. A more
complete review of the statistical data can be found in Appendix 12. Additionally, the
consequences of some of these results are discussed further in Chapter 5.
Pre-Test Findings
I distributed identical pre-test questionnaires to the control group and the
experimental group prior to the analysts beginning their research. The subjects were
asked to answer questions on a 5 level Likert scale. The purpose of the questions was to
determine if there were any significant factors that would explain differences between the
two group’s results that were not due to the experiment itself. Specifically, these
questions were designed to provide insight as to the subject’s level of interest in
participating in the experiment, as well as their level of familiarity and interest associated
with the topic (Russia’s relationship with OPEC).
37
The first question was designed to determine the level of interest regarding
participation in the experiment between the two groups. Both groups expressed roughly
the same level of interest regarding participation in the experiment, with the control
group expressing an average interest level of 3.48 (out of 5) and the experimental group
expressing an average interest level of 3.75 (out of 5). Statistical analysis revealed a P-
value of 0.586, meaning that this result was not significant. Extra credit appeared to be a
driving force behind the majority of subject participation with the control group average
at 3.9 (out of 5) and the experimental group average at 4.25 (out of 5). Specifically, extra
credit appeared to be important to the undergraduate and first year graduate students.
Only four students indicated that extra credit was “not at all important,” all of whom were
second year graduate students.
The next two questions were designed to see if one group had more knowledge in
regards to the subject matter (Russia’s relationship with OPEC), or interest in the topic,
prior to conducting research. Results indicated that the control group was more
knowledgeable on the topic, with an average score of 2.5 (out of 5), compared to the
average score of 1.7 (out of 5) expressed by the experimental group, a difference found to
be statistically significant with a P-value of 0.012. The control group also indicated that
they were slightly more interested in the topic, with an average control group score of
3.48 (out of 5) and an experimental group average of 2.83 (out of 5), a difference also
found to be statistically significant with a P-value of 0.032.
38
Post-Test Findings
As discussed in Chapter 3, the control group and experimental group were
required to fill out a post-test questionnaire upon completion of their analysis. Each
group received several identical questions. These questions were designed to provide
insight as to the frequency that each subject uses analytic methods when providing
analysis and their opinion regarding the ability of analytic methods to improve
intelligence analysis. Additionally, the questions were designed to determine the
subject’s opinion in regards to having adequate information and time to complete his or
her analysis, and to assess their overall enjoyment of the process. When I asked the
student-analysts how frequently they use structured analytic methods when solving
intelligence related problems, the control group responded with an average of 3.14 (out of
5) and the experimental group responded with an average of 3.5 (out of 5). This result
has a P-value of 0.150, thus this difference was not found to be statistically significant.
In addition, I asked the two groups to rate their level of enjoyment during the process.
The control group average was 3.7 (out of 5) and the experimental group’s average was
3.46 (out of 5). This result yielded a P-value of 0.101, thus the difference was not found
to be statistically significant. These results suggest that both the control group and the
experimental group felt that they had access to a reasonable amount of information that
was necessary to evaluate the problem to its full extent, and that the use of the
methodology by the experimental question did not seem to overly burden the student-
analysts.
I asked the student-analysts to determine if they were given enough time to
complete their analysis. The control group’s average response was 4.43 (out of 5) and
39
the experimental group responded with an average of 4.83 (out of 5), a difference that
was statistically significant with a P-value of 0.025. This result suggests that the
experimental group felt that they had more than enough time to complete the estimate, a
result that is particularly interesting when compared to the actual time each group took to
complete the estimate. (See the section on timeliness discussed later in this chapter).
Aside from the questions previously discussed, I asked both groups several
questions that were designed to gain feedback regarding the experiment set-up, as well as
gauge the student’s prior exposure to analytic methods and desire to use structured
methodology in future analysis. These questions were not tested for statistical
significance as they did not likely have any influence regarding the actual process or
product of the student’s analysis. These questions merely had to do with the experiment
itself and therefore, there was no basis for comparison.
In the post-test questionnaire, I asked the student-analysts in both groups how
familiar they were with the process of MCIM. It was predictable that the experimental
group would indicate that they were more familiar with the process of MCIM (after all
they did just receive instruction on the methodology). The control group rated a level of
familiarity as an average of 2.29 (out of 5) with six students claiming they had “no
familiarity at all with the methodology.” The experimental group rated a level of
familiarity of 3.29 (out of 5); with three students claiming they had “no familiarity at all
with the methodology.” Based on the experimental group’s responses it is likely that the
three students who indicated that they had “no familiarity at all with the methodology”
interpreted the question to mean familiarity prior to the experiment session. After all, the
40
students had just received instruction on the methodology and completed their analysis
using the methodology.
Despite the difference in level of familiarity regarding MCIM, both groups did
indicate very similar levels of familiarity with another structured analytic method, the
ACH. The control group indicated an average of 4.29 (out of 5), with no subjects
providing a response of 1 or 2 (responses that would have indicated no familiarity with
the methodology, or minimal at best). The experimental group indicated an average of
4.42 (out of 5), with no subjects provided a response of 1, 2, or 3 (responses that would
have indicated no familiarity with the methodology, or minimal at best). In spite of these
statistics indicating that the students had some level of familiarity with analytic methods,
no student-analyst in the control group actually used a structured method during their
analysis.
Product
I asked the student-analysts the following question: “How is Russia likely to seek
to interact with OPEC after the December 17th, 2008 meeting in Algeria.” Figure 4.1
expresses the COAs derived from the 21 members of the control group. Figure 4.2
expresses the COAs derived from the 24 members of the experiment group. However,
one student-analyst in the experiment group provided two answers; therefore the data in
that chart reflects 25 responses instead of 24.
41
Figure 4.1 Control Group Results
Figure 4.2 Experimental Group Results
The pie charts show that the experimental group arrived at a broader range of
possible COAs. The control group came up with five possible COAs; whereas, the
experimental group came up with seven possible COAs. Four of the COAs were roughly
the same between the two groups (Russia will closely coordinate with OPEC regarding
production, Russia will build a reserve to store oil, Russia will develop stronger ties with
OPEC, and Russia will engage in future talks with OPEC). The majority of the student-
analysts in both groups predicted that Russia will closely coordinate with OPEC
regarding oil production. This difference in number of suggested possible COAs may
imply that the use of MCIM aided the student-analysts in generating more creative and
42
unique solutions to the problems, rather than choosing the first COA that seemed logical
without reviewing alternatives.
When I originally picked the real world scenario (Russia’s relationship with
OPEC after the December 17th 2008 meeting) for the student-analysts to provide their
analysis, I chose a situation that I felt would have a clear cut answer within a few weeks
of the meeting’s termination. This has proved to not be the case up to this point. With no
clear cut answer at this time, it is impossible to evaluate the method with regard to the
accuracy of its product.
Process
Despite the lack of a clear cut answer to the question, it is beneficial to examine
the process by which both the control and experimental group derived their conclusions.
The Intelligence Community Directive (ICD) Number 203 is a document that manages
“the production and evaluation of national intelligence analysis.”51 While the entire
document does not apply to my research, several of the IC Analytic Standards outlined in
section D4 are worthy of noting. These standards include timeliness, expression of
analytic confidence, objectivity and the incorporation of alternative analysis, as well as
the use of logical argumentation.
Timeliness
The ICD recognizes the importance of timeliness in intelligence analysis. The
document emphasizes that analytic products have the obligation to meet the time 51 Intelligence Community Directive, Number 203. Analytic Standards (Effective June 21, 2007)
Available at: http://docs.google.com/gview?a=v&attid=0.1&thid=11f1d65194a3fe1b&mt=application%2Fpdf&pli=1; Internet.
43
standards set by the consumer, to ensure that they can be actionable products. This is
why I choose to evaluate the length of time it took each analyst to complete their work, as
there would be no purpose in utilizing a time consuming method; especially, if it is not
likely to yield better results.
Both groups were given two hours to complete their analysis. At the end of each
experiment section, each subject also recorded the actual time it took them to complete
their analysis, not including the instruction session. The results regarding time were
interesting. In the control group, the amount of time to complete analysis ranged from a
minimum of 45 minutes to a maximum of 105 minutes, with the average time at 70
minutes. The experimental group completed their analysis slightly faster, with a
minimum completion time of 30 minutes, a maximum completion time of 90 minutes,
and an average completion time of 58.96 minutes. The average length of completion
time was a result found to be statistically significant, with a P-value of 0.036. This result
suggests that the experimental group was able to structure and complete their analysis in
a notably more time efficient way, almost 20% faster, than the control group. Figure 4.3
represents the completion time for 20 student-analysts in the control group (one member
failed to provide their length of completion time) and for the 24 student-analysts in the
experimental group.
Figure 4.3 Length of Analysis Completion Time
44
Upon completion, the student-analysts were also asked if this was enough time to
complete their analysis. The control group indicated an average response of 4.43 (out of
5) and the experiment group indicated a response of 4.83 (out of 5), a difference that was
found to be statistically significant with a P-value of 0.025. The results regarding time
are important for two reasons. Not only were the student-analysts able to complete their
analysis more quickly, but their perception for the process was that they had more than
enough time.
Analytic Confidence
The ICD also mandates that analytic products express the analyst’s level of
confidence in their analytic judgment. After analysis was completed, each student-
analyst was asked to rate his or her level of analytic confidence.
The control group had a minimum level of analytic confidence of 1, a maximum
level of analytic confidence level of 9, and an average response of 5.93. The
45
experimental group expressed a slightly higher level of analytic confidence, with a
minimum level of 1.5, a maximum level of 9, and an average of 6.31. This difference
was not found to be statistically significant, with a P-value of 0.2335. Figure 4.4 displays
the levels of analytic confidence for both groups.
Figure 4.4 Level of Analytic Confidence
Objectivity and the Incorporation of Alternative Analysis
The ICD also mandates that analysts remain objective in their work and that their
analytic products incorporate alternative analysis when appropriate. This was one of the
primary differences I saw between the control group and the experimental group. While
a few students in the control group provided one or two alternative COAs, the majority of
the student-analysts merely provided one COA with few comparisons to any alternatives,
thus, not providing any insight to whether or not alternative solutions were considered.
In the experimental group, the student-analysts, who used MCIM, provided a list of all
possible COAs, and identified the importance of specific criterion or various factors to
46
those COAs. Their matrices highlighted strengths and weaknesses of each COA allowing
a hypothetical decision maker (or in this case, researcher) to compare each COA against
the alternatives. The matrices received from the experimental group provided a much
clearer outline regarding the thought process of each student-analyst and allowed me to
determine what alternative factors and COAs that the student-analysts considered.
Additionally, by providing a list of alternatives, I can assume that the student-analysts
took a more comprehensive look at different options, and did not merely find information
to support the first COA they found fitting.
Logical Argumentation
The ICD requires that key findings and analysis be supported by appropriate facts
and information, as well as address any divergent ideas. As part of the research process, I
asked both groups of student-analysts to use bullet points to explain their analysis.
Although the bullets provided some indication as to the thought process of the student-
analyst, the matrices completed by the experimental group were much easy to understand
and visibly conveyed the written analysis provided.
Bullet Point Analysis
I also asked the students to provide bullet point analysis so that I could capture the
data later on. One method I used to determine the thoroughness of analysis was by
evaluating the number of bullet points that each student-analyst used. I found the average
number of bullet points for each group to determine if there was a difference between
groups. The control group had a total of 146 bullet points, used a minimum of three
47
bullet points, a maximum of 11 bullet points, and an average of 6.95 bullet points. The
experimental group had a total of 139 bullet points, used a minimum of three bullet
points, a maximum of 15 bullet points and an average of 5.79 bullet points. This
difference was found to be statistically significant, with a P-value of 0.0415. I did not
evaluate the quality of each bullet point in this step; however, I do recognize that each
bullet point may not have been entirely relevant to the analysis. I am assuming that the
amount of insignificant bullet points between the two groups was roughly the same;
therefore, they would cancel each other out. Figure 4.5 provides a statistical breakdown
of the number of bullet points used by each member of the control and experimental
groups.
Figure 4.5 Bullet Point Analysis
48
Potential of Future Use
Since MCIM is relatively new to the field of intelligence, I wanted some feedback
as to the student’s opinion of future use of the methodology. Upon completion of their
analysis, I asked the experimental group several post-test questions designed to determine
if the student-analysts had enough instruction to complete the process, determine the ease
of using MCIM, and assess the likelihood that each analyst would use MCIM in the
future. When I asked the student-analysts if they felt like they had received enough
instruction to complete the process using MCIM (the group received a 35 minute lecture
on the process of MCIM in which I provided background information, and the group
worked through two example problems), they responded with an average answer of 4.5
(out of 5). Most of the student-analysts indicated this method was relatively simple to
use, rating it with an average score of 3.83 (out of 5). The majority of the student-
analysts also indicated that this was a method they would likely use again in the future,
with an average score of 3.67 (out of 5). Thus, even with minimal instruction time, the
student-analysts felt that the methodology was relatively easy to understand and use, and
it is something they would consider using in future intelligence analysis.
49
CHAPTER 5:
CONCLUSIONS
The purpose of this experiment was to test the value of using MCIM when solving
intelligence related problems. By controlling the method in which student-analysts
worked through an intelligence scenario, I was able to evaluate the significance of MCIM
in intelligence analysis. These factors included the analyst’s thought process and
product, their level of analytic confidence, as well as the length of time it took to
complete their analysis.
Pre-Test and Post-Test Conclusions
The pre-test and post-test questionnaire proved to be valuable, as they were an
indicator as to some of the factors outside of the experiment that may have contributed to
the results in some way. Additionally, they gave me a basis for which to compare the
base level of knowledge and level of interest in participation between the two groups,
prior to beginning research.
One interesting conclusion drawn from the pre-test was that although the
experimental group indicated a lower level of knowledge in regards to the topic (Russia’s
relationship to OPEC) and expressed a lower level of interest with the topic, both of
which were found to be statistically significant, they were able to arrive at a broader
range of possible COAs. This result suggests that despite the control group having a
jump start on the research, the experimental group was able to provide roughly the same
possible COAs, as well as a few alternatives. Another interesting conclusion drawn after
50
reviewing the products of the two groups occurred in regards to time. The experimental
group, who used MCIM, was able to complete their analysis in a more time efficient way
(faster), while still perceiving that they had more than enough time (according to the pre-
test). Specifically, the average completion time for the control group was 70 minutes and
the average time for experimental group was 58.6 minutes. Therefore, when looking at
the big picture, although the experimental group seemed less knowledgeable and less
interested, they were able to arrive at a more complete list of relevant possible COAs, and
they completed their analysis in less time.
Finally, the bullet point analysis indicated that the control group averaged
approximately one more bullet point in the analysis part of their product. This difference
was found to be statistically significance with a P-value of 0.0415. Although this result
suggests that the control group was able to provide more analysis than the experimental
group, it is important to remember that the worth of each bullet point was not assessed.
It is also entirely possible that the student-analysts in the experimental group felt that the
matrix was able to more accurately support their analysis, thus, they provided less bullet
point reasoning.
The results from the post-test suggest that all student-analysts participating in the
experiment had some level of familiarity with structured analytic methodologies. In
addition to familiarity with structured analytic methods, the results of the post-test
indicated that all of the subjects that participated in the experiment have actually used
these methods when conducting prior analysis. They also expressed that these methods
improve intelligence analysis on some level. Despite these factors, no one in the control
group utilized any sort of structured method when completing their analysis. There is no
51
apparent reason as to why this happened. The student-analysts may have felt that
structured methodology wasn’t necessary for this type of problem, or perhaps they felt
that using an informal process was both quicker and easier. Any future research or
experiments conducted regarding the use of structured methodology in intelligence
analysis, should include a post-test question asking why the student either chose to use, or
chose not to use methodology. By asking this question, more information could be
gathered as to what makes an analyst chose the method that they did.
Process and Product Conclusions
As previously mentioned, when I originally picked the real world scenario
(Russia’s relationship with OPEC after the December 17th, 2008 meeting) for which the
student-analysts to provide their analysis, I chose a situation that I felt would have a clear
cut answer within a few weeks. Although this has proved to not be the case, the use of a
MCIM model has proven to be beneficial in a variety of ways. The matrix clearly
explains how each analyst in the experimental group arrived at their conclusion. The
student-analysts outlined the criteria that he or she felt to be important and explained how
significant each of these factors was at arriving at their decision by the use of a weighting
system. Additionally, if any of the criteria should change as time progresses, the matrix
is easily modified by adding or eliminating criteria, or by merely adjusting the weights to
arrive at an alternative COA. In other words, all the hard work has already been done on
the problem, and only minor changes would need to be made to suggest an alternative
COA needed due to changes in the adversary’s environment or priorities. Since no one in
the control group used any form of structured methodology to arrive at their conclusion,
52
their decision maker (or in this case, researcher) would have a minimal idea of how each
arrived at their conclusion. Although a majority of the control group’s bullet point
analysis does support the COA that they chose to be most likely, there is minimal
mention of any other possible COA, or of the significance of each factor they considered
to support their predicted COA. If there was a change in the adversary’s environment
that would suggest the need for an alternative COA, the student-analysts in the control
group would need to go back and figure out what is the next most likely solution based
on the importance of each factor that they considered.
In general MCIM is beneficial to the field of intelligence due to its flexibility and
versatile nature. The use of MCIM is not specific to a given intelligence field, therefore
it is applicable to business, law enforcement and national security. Additionally, it is an
appropriate method to use when working with both qualitative and quantitative data.
A Step Forward
While there are a number of potential benefits with using this methodology, there
are also several prospective weaknesses. Although the matrix seeks to identify the most
important criteria involved in the decision making, it has the potential to be flawed as the
analyst is doing the scoring of the criteria. Therefore, the process of assigning weight to
pieces of criteria has the potential to be based on the analyst’s intuition and may contain
bias. Additionally, an analyst must know how and when to make a distinction when
weighing different criteria. For example, I would like to refer back to the example of
buying a car that I discussed in Chapter 2. How would an analyst rank a difference in
price for values that are close, such as $25,000 and $25,001? In other words, at what
53
point do they numbers quit being the same? Finally, this process seeks to forecast
something in which the analyst has no actual control over. Individuals only have total
control over what they do, but not competitors, enemies, or criminals. Thus, even with a
detailed, structured approach that hopes to identify the COA that an opponent is most
likely to chose, it is possible that the adversary will go with a COA that is completely
random based on their intuition or “gut feeling.” Heuer draws attention to this idea when
he said, “Caution is in order, however, whenever one thinks of predicting or even
explaining another person’s decision, regardless of whether the person is American or
foreign. People do not always act rationally in their own best interests. Their decisions
are influenced by emotions, habits, what others might think, or values that others many
not be aware of.”52
In all, the benefits of using a structured methodology in intelligence analysis, as
well as the overall general positive responses received from the student-analysts
regarding the use of MCIM, suggest that this topic is worthy of continued research.
Larger sample sizes would be beneficial in any future experiments in order to confirm
hypotheses surrounding use of the method. Additionally, in future experiments, analysis
should be provided for a real world situation that would have a clear cut answer within a
short time frame of the analysis in order to further evaluate MCIM’s ability to aid
decision makers in efficient forecasting.
Despite the possible flaws with the methodology, the results of this experimental
research highlight the potential benefits of using structured analytic methods, such as
MCIM, during the analytic process. Although the use of MCIM can never guarantee a
“correct” answer, it may increase the odds of choosing the most likely answer, by
52 Heuer, Richards, J., Email. February 12, 2009.
54
providing a more efficient way of handling unstructured data in both qualitative and
quantitative form. Additionally, matrix analysis provides a clear explanation to a
decision maker regarding the thought process of each analyst. By identifying the weight
that an analyst places on each individual criterion, MCIM can highlight the areas of
disagreement between two individuals, and it is easily adjusted to accommodate new data
or “what-if” scenarios. Additionally, as demonstrated in the experiment, using the
MCIM methodology may allow an analyst to structure and complete their analysis in a
more time efficient.
While the jury is still out regarding the accuracy of the analysis, MCIM can help
teach an analyst the importance of considering the significance of certain factors. This
study hopes to take the first steps of many in increasing an analyst’s ability to thoroughly
examine all possible COAs in a timely manner, as well as accurately forecast what COA
one’s adversary is likely to choose.
55
BIBLIOGRAPHY
Balcomb, J.D. and A. Curtner. “Multi-Criteria Decision-Making Process for Building.” Paper to be presented at the American Institute of Aeronautics and Astronautics Conference, Las Vegas, Nevada, July 24-28. 2000. Available online at http://www.nrel.gov/docs/fy00osti/28533.pdf; Internet; accessed 23 March 2009.
Belton, Valerie and Theodor J. Stewart. Multiple Criteria Decision Analysis: An Integrated Approach. The Netherlands: Kluwer Academic Publishers, 2002.
Dzubow, PhD and William Z. Suplee IV, CFA, CFP, ChFC, CASL. “Using Multiple-Criteria Decision Analysis to Simplify the Financial Planning Process.” Journal of Financial Planning. March 2008. Available at http://www.library.idsc.gov.eg/GUI/Globals/Upload/BULLETIN_ATTACHMENT/92/e-files/manegement%20and%20economics/using%20multiple.pdf; Internet; accessed 5 December 2008.
Forgang, William G., Strategy-Specific Decsion Making: A Guide for ExecutingCompetitive Strategy. M.E. Sharpe, Inc., Armonk, New York, 2004.
Forman, Ernest, Decision by Objectives (How to Convince Others That You Are Right). Available at http://mdm.gwu.edu/forman/DBO.pdf; Internet.
Gilchrist, Stacy. “Staff Study.” Research Paper for Advanced Analytic Techniques, Mercyhurst College.
Heuer, Richards, Jr. Email. 12 February 2009.
Heuer, Richards, Jr. Psychology of Intelligence Analysis [book on-line]. Langley, Virginia: Central Intelligence Agency Center for the Study of Intelligence, 1999. Available at https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/PsychofIntelNew.pdf; Internet.
Heuer, Richards, Jr. “Taxonomy of Structured Analytic Techniques.” Paper preparedfor the International Studies Association 2008 Annual Convention, March 26-29, 2008, San Francisco, CA. Available at http://www.allacademic.com//meta/p_mla_apa_research_citation/2/5/4/1/2/pages254125/p254125-1.php; Internet; accessed 20 November 2008.
Intelligence Community Directive, Number 203. Analytic Standards (Effective 21 June 2007). Available at http://docs.google.com/gview?a=v&attid=0.1&thid=11f1d65194a3fe1b&mt=application%2Fpdf&pli=1; Internet.
56
Kiker, Gregory A., Todd S. Bridges, Arun Varghese, Thomas P. Seager and Igor Linkov. “Application of Multicriteria Decision Analysis in Environmental Decision Making.” Integrated Environmental Assessment and Management. Volume 1, Number 2, Pages 95-108. 2005. Available at http://www.allenpress.com/pdf/ieam-01-02_95_108.pdf; Internet; accessed 20 November 2008.
LeGault, Michael R. Think: Why Crucial Decision Can’t Be Made in the Blink of an Eye. New York, Threshold Editors, 2006. Page 305.
Marrin, Stephen. “Intelligence Analysis: Structured Methods or Intuition.” American Intelligence Journal. Page 7. Summer 2007.
Peterson, Joshua. “Appropriate Factors to Consider When Assessing Analytic Confidence in Intelligence Analysis.” (Master's Thesis, Mercyhurst College, March 2008).
Satty, Thomas. “Decision making with the analytic hierarchy process,” Int. J. Services Sciences, Vol. 1, No. 1, 2008. Available at http://inderscience.metapress.com/media/pgwf2qtuyg3xnvnhxnby/contributions/0/2/t/6/02t637305v6g65n8.pdf; Internet; accessed 26 March 2009.
Saaty, Thomas L. “How To Make A Decision: The Analytic Hierarchy Process.” Available at http://sigma.poligran.edu.co/politecnico/apoyo/Decisiones/curso/Interfaces.pdf; Internet; accessed 24 March 2009.
Simon, Herbert and Associates. “Decision Making and Problem Solving.” Reprinted with permission from Research Briefings 1986: Report of the Research Briefing Panel on Decision Making and Problem Solving. 1986 by the National Academy of Sciences. Published by National Academy Press, Washington, DC. Available at http://dieoff.org/page163.htm; Internet; accessed 20 November 2008.
The United States Army. Field Manual 101-5. Staff Organization and Operations. 31 May 1997. Available at http://www.fs.fed.us/fire/doctrine/genesis_and_evolution/source_materials/FM-101-5_staff_organization_and_operations.pdf; Internet. Page 115.
The United States Army 2008 U.S. Army Posture Statement. Information Papers: Red Team Education and Training. Available at http://www.army.mil/aps/08/information_papers/prepare/Red_Team_Education_and_Training.html; Internet; accessed 26 September 2008.
The United States Government. Director of National Intelligence, Vision 2015. Available at http://www.dni.gov/Vision 2015.pdf; Internet.
57
Triantaphyllou, Evangelos and Stuart H. Mann. “An Examination of the Effectiveness of Multi-Criteria Decision Making Methods: A Decision Making Paradox.” Available at http://www.csc.lsu.edu/trianta/index.html?http://www.csc.lsu.edu/trianta/Books/DecisionMaking1/Book1.htm; Internet.
Triantaphyllou, Evangelos. “Can We Always Determine the Right Alternatives in Business Problems.” 18 August 2002.
Triantaphyllou, Evangelos and Panos M. Parlos (ed.). Multi-Criteria Decision Making Methods: A Comparative Study. The Netherlands: Kluwer Academic Publishers, 2000.
Wang, Xiaoting. “Study of Ranking Irregularities When Evaluating Alternatives By Using Some ELECTRE Methods and a Proposed New MCDM Method Based on Regret and Rejoicing,” A Thesis Submitted to the Graduate Faculty of the Louisiana State University and Mechanical College. August 2007. Available online at http://etd.lsu.edu/docs/available/etd-07112007-012708/unrestricted/Wang_thesis.pdf; Internet; accessed 24 March 2009.
Webster’s Dictionary Online; available from http://www.merriamwebster.com/dictionary/criteria; Internet.
Zopounidis, Constantin and Doumpos, Michael. “Multiple-Criteria Decision Making.” Thomson Corporation. 2006.
58
APPENDICES
59
Appendix 1:
Analytic Methods Experiment Sign-Up Form
Name:
Class Year:
Phone Number:
E-mail Address:
Instruction Session Dates/Times:(Please select the day you are able to participate)
Wednesday, 10 December 2008 5:00-8:00pm ____
Thursday, 11 December 2008 5:00-8:00pm ____
If you are able to participate on either day please ____check here, and you will be randomly assigned a day
*** I will send you an e-mail confirming your experiment day and time
Upon completion, please return this form to Lindsey Jakubchak or Travis Senor in CIRAT.
Contact Info:[email protected]
60
Appendix 2:
Institutional Review Board Proposal Form
Date Submitted:10/10/2008
Investigator(s):Lindsey Jakubchak
Investigator Address:
Investigator(s) E-mail:[email protected]
Investigator Telephone Number:
Advisor's Name (if applicable):Kristan Wheaton
Advisor’s E-mail:[email protected]
Advisor's Signature of Approval:[X] Place X here if advisor hasapproved research
Title of Research Project:The Effectiveness of Multi-Criteria Decision Making (MCDM) in the Field of Intelligence
Date of Initial Data Collection:TBD, anticipate October-December 2008
________________________________________________________________________
Please describe the proposed research and its purpose, in narrative form:
Multi-Criteria Decision Making (MCDM) is a generic term used to encompass a broad range of analytic methodologies that use matrices as the basis for their conclusions. More specifically, MCDM is an internally focused support system that confronts real world decisions based on the evaluation of multiple criteria, goals, and objectives of conflicting nature. Additionally, it provides substance to a decision maker as it allows for selecting an option based on the appropriateness of those alternatives weighted against the others.
While there is a substantial body of literature related to the use and effectiveness of MCDM in the general sense, there has been little research done in the field of intelligence. The purpose of this proposed research and experiment is to take an alternative look at this methodology. In other words, if MCDM was switched to an external focus, and used in the field of intelligence, is it likely that this methodology
61
would be a valuable way to predict what course of action one’s adversary is likely to choose?
I have developed an experiment which I feel will test this hypothesis, in both process and product, when solving intelligence related problems.
I plan to test if MCDM truly is a viable methodology by using intelligence analysts (both undergraduate and graduate students in the intelligence program at Mercyhurst College) to solve a real world problem.
Indicate the materials, techniques, and procedures to be used (submit copies of materials):Materials:
Exercise ScenariosWriting UtensilsPre-Test and Post-Test Questionnaire
Procedure:
One month prior to the experiment, I will solicit both undergraduate and graduate students at Mercyhurst College to participate through department wide e-mails, fliers and through short information sessions at the beginning of classes (with permission of the designated professor). Students will be selected on a first come, first serve basis. One week prior to the study, I will send out reminders to those who have volunteered to participate. I will send out another reminder the day before the study.
During the actual study, I will begin by going over the directions. I will then explain what is expected of the participants, and how they will be getting credit for their participation. After the introduction, I will pass out a pre-test questionnaire (attached at the end of this form), the study’s materials, and instruct the students to begin their analyses. I will then provide the students with one week in which they can complete their analysis and provide a time in which they can return their finished product to me. This process will be the same for both groups (the control group and the experimental group), with the only exception being that the experimental group will receive a one hour long instructional session on the how to use a variant of MCDM in an intelligence context (in order to complete their analysis), after the introduction.
Following the completion of the exercise, I will ask the participants to fill out a questionnaire (attached at end of this form) and provide feedback regarding both the topic and the experiment.
I plan to conduct my experiment two times, on two different nights. They will vary only in which handout I give them. There will be one experimental group and a control group.
62
All groups will be given the exact same scenario; I will just vary the methodology utilized to solve the problem. (Please see the forms below)
63
1. Do you have external funding for this research (money coming from outside the College)? Yes[ ] No[X]
Funding Source (if applicable):
2. Will the participants in your study come from a population requiring special protection; in other words, are your subjects someone other than Mercyhurst College students (i.e., children 17-years-old or younger, elderly, criminals, welfare recipients, persons with disabilities, NCAA athletes)? Yes[ ] No[X]
If your participants include a population requiring special protection, describe how you will obtain consent from their legal guardians and/or from them directly to insure their full and free consent to participate.
N/A
Indicate the approximate number of participants, the source of the participant pool, and recruitment procedures for your research:
I plan to have approximately 75 participants. I plan to recruit undergraduate and graduate students in the intelligence studies department through a department-wide email, fliers hung in the building, and by information sessions held at the beginning of classes (with permission of the designated professor). I will select the students on a first come, first serve basis.
Will participants receive any payment or compensation for their participation in your research (this includes money, gifts, extra credit, etc.)? Yes[X] No[ ]
If yes, please explain:
In the past, most of the intelligence professors have been willing to grant extra credit for participating in an experiment.
3. Will the participants in your study be at any physical or psychological risk (risk is defined as any procedure that is invasive to the body, such as injections or drawing blood; any procedure that may cause undue fatigue; any procedure that may be of a sensitive nature, such as asking questions about sexual behaviors or practices) such that participants could be emotionally or mentally upset? Yes[ ] No[X]
Describe any harmful effects and/or risks to the participants' health, safety, and emotional or social well being, incurred as a result of participating in this research, and how you will insure that these risks will be mitigated:
None.
64
4. Will the participants in your study be deceived in any way while participating in this research? Yes[ ] No[X]
If your research makes use of any deception of the respondents, state what otheralternative (e.g., non-deceptive) procedures were considered and why they weren't chosen:
N/A
5. Will you have a written informed consent form for participants to sign, and will you have appropriate debriefing arrangements in place? Yes[X] No[ ]
Describe how participants will be clearly and completely informed of the true nature and purpose of the research, whether deception is involved or not (submit informed consent form and debriefing statement):
Prior to the training sessions, participants will be provided with a general overview of what will occur during the session as well as the consent form, which will also describe what is expected of them. Following the administrative questionnaire, participants will be provided with a debriefing statement that will explain how the results from the session will be used (please see forms at the end of this proposal).
Please include the following statement at the bottom of your informed consent form: “Research at Mercyhurst College which involves human participants is overseen by the Institutional Review Board. Questions or problems regarding your rights as a participant should be addressed to Mr. Tim Harvey, Institutional Review Board Chair; Mercyhurst College; 501 East 38th Street; Erie, Pennsylvania 16546-0001; Telephone (814) 824-3372.”
6. Describe the nature of the data you will collect and your procedures for insuring that confidentiality is maintained, both in the record keeping and presentation of this data:
Names are not required for my research and thus no names will be used in the recording of the results or the presentation of my data. Names will only be used to notify professors of participation in order for them to correctly assign extra credit.
7. Identify the potential benefits of this research on research participants and humankind in general.
Potential benefits include:
For participants:For some students, this experiment will provide an opportunity to learn a new analytical tool that they may use in intelligence related decision making. For other students, it will provide an opportunity to practice decision making skills that they have learned in the classroom or developed in the real world. Students are often asked to utilize various
65
analytical tools in order to gain a more thorough understanding of the problem at hand, and this may be an effective methodology for doing that.
For the Intelligence Community (IC):This experiment hopes to prove that MCDM is an effective methodology to use when making intelligence related decisions. In doing so, the IC would benefit from a methodology that allows our nation’s decision makers to make more informed decisions.
Please submit this file and accompanying materials to the IRB Chair, Tim Harvey, via electronic mail ([email protected]) for review.
Appendix 3:
Multi-Criteria Decision Making Participation Consent Form
The purpose of this research is to test the effectiveness of a particular method, in both process and product, when making intelligence related decisions. Your participation involves an instruction period, completion of a short intelligence analysis, and filling out a pre and post test questionnaire. This process should take no longer than 3 hours. Your name WILL NOT appear in any information disseminated by the researcher. Your name will only be used to notify professors of your participation in order for them to assign extra credit.
There are no foreseeable risks or discomforts associated with your participation in this study. Participation is voluntary and you have the right to opt out of the study at any time for any reason without penalty.
I, ____________________________, acknowledge that my involvement in this research is voluntary and agree to submit my data for the purpose of this research.
_________________________________ __________________Signature Date_________________________________ __________________Printed Name Class
Name(s) of professors offering extra credit and class(es) you are enrolled in: (Professor Breckenridge, Professor Marrin, Professor Welch, and Professor Wheaton) ________________________________________________________________________
Researcher’s Signature: __________________________________________________________
If you have any further question about methodology used, or this research, you can contact me at [email protected]
Research at Mercyhurst College which involves human participants is overseen by the Institutional Review Board. Questions or problems regarding your rights as a participant should be addressed to Tim Harvey; Institutional Review Board Chair; Mercyhurst College; 501 East 38th Street; Erie, Pennsylvania 16546-0001; Telephone (814) 824-3372. [email protected]
Lindsey Jakubchak, Applied Intelligence Master’s Student, Mercyhurst College
Kristan Wheaton, Thesis Advisor, Mercyhurst College
Appendix 4:
Experiment Section #1 (Control Group)
You are a national security analyst specializing in Russian relations. Your director is interested in the outcome of the upcoming meeting of the Organization of Petroleum Exporting Countries (OPEC), taking place in Algeria on December 17th, 2008. Please provide an answer to the following question:
How is Russia likely to seek to interact with OPEC after the December 17th, 2008 meeting in Algeria?
STEP 1: Please complete the following analysis in the space provided below. *You may use any sources you would like
It is likely that Russia will seek to-________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
STEP 2: Please list your reason(s) for your analysis: *You do not need to provide your sources
STEP 3: Please rate your level of analytic confidence.
Important Information:Analytic Confidence:
Analytic Confidence reflects the level of confidence an analyst has in his or her estimates and analyses. It is not the same as using words of estimative probability, which indicate likelihood. It is possible for an analyst to suggest an event is virtually certain based on the available evidence, yet have a low amount of confidence in that forecast due to a variety of factors or vice versa.
As you are considering your level of analytic confidence, please consider these seven factors:
1. Use of Structured Method(s) in Analysis2. Overall Source Reliability3. Source Corroboration/Agreement4. Level of Expertise on Subject/Topic & Experience5. Amount of Collaboration6. Task Complexity7. Time Pressure
Mark your analytic confidence on the scale below.
Low |-----------------------------------------------------------------------| High
Appendix 5:
Analytic Methods Experiment Links
Current Situation: Although Russia is not currently an OPEC member, it does maintain a
relationship with the organization.
Links to get you started… You are not limited to these sources; they are merely starting points to familiarize
yourself with the topic! Feel free to use whatever sources you would like!
Monthly Oil Market Report-November 2008:Source: Organization of the Petroleum Exporting Countries (OPEC)http://www.opec.org/home/Monthly%20Oil%20Market%20Reports/2008/pdf/MR112008.pdf
Non-OPEC Oil Production:Source: Council on Foreign Relationshttp://www.cfr.org/publication/14554/nonopec_oil_production.html
Organization of the Petroleum Exporting Countries (OPEC):Source: OPEC Homepagehttp://www.opec.org/home/
Outlook for Non‐OPEC Oil Supply Growth in 2008‐2009:Source: Energy Information Administrationhttp://www.eia.doe.gov/emeu/steo/pub/special/2008-non-opec-oil-supply.pdf
Russia Energy Data, Statistics and Analysis - Oil, Gas, Electricity, Coal:Source: Energy Information Administrationhttp://www.eia.doe.gov/emeu/cabs/Russia/pdf.pdf
Appendix 6:
Pre-Test Questionnaire (Control Group and Experimental Group)
Answer the following questions by circling the best response. Please answer all the questions honestly.
On a scale from 1 to 5, where 1 represents not interested at all and 5 represents extremely interested, how interested are you in participating in this experiment?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not knowledgeable at all and 5 represents extremely knowledgeable, how knowledgeable are you with this topic?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not interested at all and 5 represents extremely interested, how interested are you in this topic?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not interested at all and 5 represents extremely interested, how interested are you in learning about analytic methodologies?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough time and 5 represents more than an adequate amount of time, do you think you will have enough time to complete this analysis?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not important at all and 5 represents extremely important, how important was offering extra credit in participating in this experiment?
1 2 3 4 5
What is your class rank?
FR SO JR SR G1 G2
Please provide any additional comments or feedback in the space provided:________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
Appendix 7:
Post-Test Questionnaire (Control Group)
Answer the following questions by circling the best response. Please answer all the questions honestly.
On a scale from 1 to 5, where 1 represents not at all familiar and 5 represents extremely familiar, how familiar are you with the process of Multi-Criteria Decision Making (MCDM)?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all familiar and 5 represents extremely familiar, how familiar are you with the process of Analysis of Competing Hypothesis (ACH)?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all and 5 represents extremely frequently, how frequently do you use structured analytic methods in solving intelligence related problems?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents no improvement at all and 5 represent a significant improvement, how much do you feel analytic methods improve intelligence analysis?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough at all, and 5 represents more than enough, do you feel you found enough adequate information to complete this analysis?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough time at all and 5 represents more than adequate amount of time, do you feel you had enough time to complete this process?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents no enjoyment at all and 5 represents thorough enjoyment, how much did you enjoy this process?
1 2 3 4 5
Approximately how much time did you spend completing this analysis?___________________
What do you think the purpose of this experiment was?________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
What is your class rank?
FR SO JR SR G1 G2
Please provide any additional comments or feedback in the space provided:________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
Appendix 8:
Multi-Criteria Decision Making Participation Debriefing
Thank you for participating in this research process. I appreciate your contribution and willingness to support the student research process.
The purpose of this study was to determine the effectiveness of using MCDM as an analytical methodology in the field of intelligence. Currently there has been little research done on this topic, and this study hopes to take the first of many steps in increasing one’s ability to accurately predict both what course of action one’s adversary is likely to choose, as well as forecast what is happening in the environment in which an analyst works in. My experiments today were designed to demonstrate the significance of using this methodology in solving intelligence related problems.
By shifting the process of MCDM from an internal focus to an external focus, it is keeping with the theme of intelligence. Additionally, utilizing methods that promote a more efficient method of structuring data may increase an analyst’s ability to learn what is happening with his/her adversary or the environment in which they are working in. I plan to use the results from this study to determine if the use of MCDM in solving intelligence related problems improves the analytic process and product.
Please note, you will be able to review the results of this experiment. The results will be included in a paper that I am preparing to present at the International Studies Association (ISA) Conference, and will be accessible online. Additionally, results and conclusions will be accessible via my completed thesis.
If you have any further question about MCDM or this research you can contact me at [email protected]
Appendix 9:
Multi-Criteria Decision Making Lecture
Everybody Makes Decisions On A Daily Basiso Some decisions can be made with our intuition or “gut feelings” as they
are not life altering and require little to no analysis. Examples: What will I eat for breakfast in the morning? What road
will I use to travel to work?o Other decisions require the balancing of multiple factors or criteria, and do
affect our life in some way. Examples: Where will I go to college? What kind of car will I
buy? What career will I choose to embark on? o In the field of intelligence, decision related to national security, law
enforcement, and business directly relate to the safety of a nation, the rise and fall of a country, and the stability of an organization. These decisions mandate appropriate analysis.
Examples: What is likely to happen in Country X in the next two years? What will likely happen to the oil fields in Country Y over the next 12-24 months?
Multi-Criteria Decision Making (MCDM)o The process of evaluating possible courses of action (COA) in a
systematic way against a set of fixed criteria.o Involves the breakdown of complex problems into smaller pieces in order
for more thorough analysis.o Often called “matrix analysis.”o MCDM is a generic term used to encompass a broad range of analytical
methodologies o In MCDM, emphasis is on placing value, or judgment of an item’s worth
and desirability, on pieces of criterion, a standard by which a judgment or decision may be made.
What Constitutes a MCDM problem?o MCDM is used when a decision needs to be made that has two or more
known COAs or two or more possible COAs.
o Either a “need to find best solution” or “we need to find a solution.”o Must have multiple criteria.
Process Of MCDM (8 Steps)o Requirement/Question and Collection
Good Essential Question Understanding INTENT of question, is essential to good analysis
o Establish Possible COA
Brainstorm Want to identify as many COAs as possible
o Establish “Screening Criteria” Designed to eliminate COAs “Must have/be” guidelines
o Screen COAo Establish “Evaluation Criteria”
Designed to rank COAs Can be same as screening criteria.
o Weight Evaluation Criteriao Evaluate COA
What scale? Best to worst, 1 to 3, 1 to 5, 1 to 10? Based on? Intuition? Standards? Even distribution?
o Make Recommendations/Estimate
Conventional Use Of MCDM Vs. Use Of MCDM In Intelligenceo Difference of perspective
The conventional MCDM process focuses on one’s own decision-making process, a situation where an individual has complete control.
The intelligence focused MCDM process focuses on the decision-making process of others, a situation in which an analyst has no control over.
o Purpose By shifting the process of MCDM from an internal focus to an
external focus, it is keeping with the theme of intelligence Therefore, instead of establishing the COAs about the “best”
decision for an individual, an analyst seeks to determine which COA one’s adversary or competitor is likely to choose.
Likely Benefits Of The MCDM Process In Intelligenceo Utilizing methods that promote a structured review of external information
increases an analyst’s awareness of his/her adversary, and the environment in which they are working in.
By continually assessing the social, political, economical and technological aspects of an adversary’s environment, an analyst might gain insight to factors outside of a decision maker’s immediate control, deterring them from decisions they would otherwise likely make.
o The use of an intelligence focused MCDM process may help to prevent analytic pitfalls, such as the satisficing strategy and groupthink.
o Should allow for a more complete analysis of the situation and account for possible COAs that would otherwise have not been considered.
Appendix 10:
Experiment Section #2 (Experimental Group)
You are a national security analyst specializing in Russian relations. Your director is interested in the outcome of the upcoming meeting of the Organization of Petroleum Exporting Countries (OPEC), taking place in Algeria on December 17th, 2008. Using Multi-Criteria Decision Analysis, please supply an answer to the following question:
How is Russia likely to seek to interact with OPEC after the December 17th, 2008 meeting in Algeria?
The following is designed to guide you through the process of MCDM. Please note, Steps 2, 3 and 4 are mandatory. Step 1 is optional, and is designed to assist you with completion of the matrix. *You may use any sources you would like.
STEP 1: This step is optional
Possible Course of Action: What courses of actions are possible for Russia? You may have as many courses of action as you would like
COA 1:COA 2:COA 3:
Screening Criteria: Designed to eliminate COA’soo
Remaining Courses of Action: What courses of actions are possible for Russia? (After screening applied) You may have as many courses of action as you would like
COA 1:COA 2:
Evaluation Criteria: Designed to rank COAs. Evaluation Criteria may be tangible items (things that can be measured) or intangible items (things things that cannot be measured). These criteria can be ranked by scale: best to worst 1 to 3, 1 to 5, etc. They can also be based on intuition, standards, even distribution, etc.
Eval 1:Eval 2:Eval 3:
Step 2: This step is mandatory
Matrix: Please complete the Matrix in Excel (format is on the shared drive). Below is to demonstrate what the matrix should look like.
Criteria 1: Criteria 2: Criteria 3: Criteria 4: Total/Rank Order
COA 1:
COA 2:
COA 3:
COA 4:
COA 5:
It is likely that Russia will seek to ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
STEP 3: This step is mandatory Please list your reason(s) for your analysis: *You do not need to provide your sources
STEP 4: This step is mandatory
Please rate your level of analytic confidence.
Important Information:Analytic Confidence:
Analytic Confidence reflects the level of confidence an analyst has in his or her estimates and analyses. It is not the same as using words of estimative probability, which indicate likelihood. It is possible for an analyst to suggest an event is virtually certain based on the available evidence, yet have a low amount of confidence in that forecast due to a variety of factors or vice versa.
As you are considering your level of analytic confidence, please consider these seven factors:
1. Use of Structured Method(s) in Analysis2. Overall Source Reliability3. Source Corroboration/Agreement4. Level of Expertise on Subject/Topic & Experience5. Amount of Collaboration6. Task Complexity7. Time Pressure.
Mark your analytic confidence on the scale below.
Low |-----------------------------------------------------------------------| High
Appendix 11:
Post-Test Questionnaire (Experimental Group)
Answer the following questions by circling the best response. Please answer all the questions honestly.
On a scale from 1 to 5, where 1 represents not at all familiar and 5 represents extremely familiar, how familiar are you with the process of Multi-Criteria Decision Making (MCDM)?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all familiar and 5 represents extremely familiar, how familiar are you with the process of Analysis of Competing Hypothesis (ACH)?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all and 5 represents extremely frequently, how frequently do you use analytic methods in solving intelligence related problems?
1 2 3 4 5
On a scale from 1 to 5, where 1 no improvement and 5 represents a significant improvement, how much did this using this methodology improve your intelligence analysis?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents no enjoyment at all and 5 represents thorough enjoyment, how much did you enjoy this process?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough at all, and 5 represents more than enough, do you feel you found enough adequate information to complete this analysis?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough time at all and 5 represents more than adequate amount of time, do you feel you had enough time to complete this process?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not enough instruction at all and 5 represents more than enough instruction, do you feel you had enough instruction to complete this process using MCDM?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents extremely difficult and 5 represents extremely easy, how was this methodology to use?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all interested and 5 represents extremely interested, how interested are you in learning about different analytic methods after completion of this exercise?
1 2 3 4 5
On a scale from 1 to 5, where 1 represents not at all likely and 5 represents extremely likely, how likely is it that you will use this methodology in the future?
1 2 3 4 5
Approximately how much time did you spend completing this analysis (Not including classroom instruction)?___________________
What do you think the purpose of this experiment was?________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
What is your class rank?
FR SO JR SR G1 G2
Please provide any additional comments or feedback in the space provided:______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________-_____________________________________________________________________________
Tests of Normality
.221 20 .012 .924 20 .116
.212 20 .019 .935 20 .193
Group A (Control)
Group B (Experiment)
Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Appendix 12:
Statistical Data
Length of Time to Complete Analysis (in minutes):
Null: Group B does not take less time to finish.
Alternative: Group B takes less time to finish than group A. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Shapiro-Wilk test gives p-values > ( = 0.05), thus normality assumption is satisfied for both samples.
110100908070605040
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
al
Normal Q-Q Plot of Group A (Control)
Most points are close to the line thus the assumption of normality is satisfied for the group A.
90807060504030
Observed Value
2
1
0
-1Exp
ecte
d N
orm
al
Normal Q-Q Plot of Group B(Experiment)
Group A (Control)
110
100
90
80
70
60
50
40
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Box plot shows no outliers for group A.
Group Statistics
20 70.0000 17.09109 3.82168
24 58.9583 16.61450 3.39142
GroupGroup A - Control
Group B - Experiment
Length of time in minutesN Mean Std. Deviation
Std. ErrorMean
Independent Samples Test
.542 .466 2.167 42 .036 11.04167 5.09607 .75738 21.32596
2.161 40.143 .037 11.04167 5.10950 .71612 21.36721
Equal variancesassumed
Equal variancesnot assumed
Length of time in minutesF Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)Mean
DifferenceStd. ErrorDifference Lower Upper
95% ConfidenceInterval of the
Difference
t-test for Equality of Means
Independent Samples Test
.542 .466 2.167 42 .036
2.161 40.143 .037
Equal variancesassumed
Equal variancesnot assumed
Length of time in minutesF Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Group B (Experiment)
90
80
70
60
50
40
30
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.466) > ( = 0.05), thus assumption of equal variances is satisfied.
Box plot shows no outliers for group B.
Tests of Normality
.165 21 .137 .945 21 .268
.137 24 .200* .939 24 .157
Groups for AnalyticConfidenceGroup A (Control)
Group B (Experiment)
Level of AnalyticConfidence
Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
This is a lower bound of the true significance.*.
Lilliefors Significance Correctiona.
According to above table, t-test value = 2.167, P-value = 0.036.
Since (P-value = 0.036) < ( = 0.05), null hypothesis is rejected.
Conclusion: At 5% level, Group B takes less time to finish than group A.
Level of Analytic Confidence:
Null: Group B does not have higher level of analytic confidence.
Alternative: Group B does have higher level of analytic confidence than group A. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Kolmogorov-Smirvov test gives p-values > ( = 0.05), thus normality assumption is satisfied for both samples.
1086420
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
alfor groupAC= Group A (Control)
Normal Q-Q Plot of Level of AnalyticConfidence
108642
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
al
for groupAC= Group B (Experiment)
Normal Q-Q Plot of Level of AnalyticConfidence
Most points are close to the line thus the assumption of normality is satisfied for the group A.
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Group Statistics
21 5.9286 1.75560 .38310
24 6.3125 1.74961 .35714
Groups for AnalyticConfidenceGroup A (Control)
Group B (Experiment)
Level of AnalyticConfidence
N Mean Std. DeviationStd. Error
Mean
Independent Samples Test
.002 .967 -.733 43 .467 -.38393 .52363 -1.43993 .67207
-.733 42.171 .468 -.38393 .52375 -1.44078 .67292
Equal variancesassumed
Equal variancesnot assumed
Level of AnalyticConfidence
F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)Mean
DifferenceStd. ErrorDifference Lower Upper
95% ConfidenceInterval of the
Difference
t-test for Equality of Means
Independent Samples Test
.002 .967 -.733 43 .467
-.733 42.171 .468
Equal variancesassumed
Equal variancesnot assumed
Level of AnalyticConfidence
F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Group B(Experiment)
Group A (Control)
Groups for Analytic Confidence
10.00
8.00
6.00
4.00
2.00
0.00
Lev
el o
f A
nal
ytic
Co
nfi
den
ce
314
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.967) > ( = 0.05), thus assumption of equal variances is satisfied.
Box plot shows outliers for both groups. Even with the presence of outliers, normality is satisfied. Also these values are important for the analysis. Thus the decision is not to remove the outliers.
Tests of Normality
.208 21 .019 .941 21 .229
.317 24 .000 .749 24 .000
Groups for BulletPoint ReasoningGroup A - Control
Group B - Experiment
Bullet Point ReasoningStatistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
According to above table, t-test value = -0.733, P-value = 0.467/2 = 0.2335 (need to divide the P-value by 2 as we have one-tailed test where as SPSS gives you 2-tailed test P-value.
Since (P-value = 0.2335) > ( = 0.05), null hypothesis is not rejected.
Conclusion: At 5% level, Group B does not have higher level of analytic confidence than group A.
Bullet Point Reasoning:
Null: Group B does not have less bullet points than group A.
Alternative: Group B does have less bullet points than group A. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Shapiro-Wilk test gives p-value > ( = 0.05), thus normality assumption is satisfied for Group A.
12108642
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
alfor groupBullet= Group A - Control
Normal Q-Q Plot of Bullet PointReasoning
Both Kolmogorov-Smirnov and Shapiro-Wilk test give p-value < ( = 0.05), thus normality assumption is not satisfied for Group B. Need to check Normal Probability plot for group B.
15129630
Observed Value
2
1
0
-1
Exp
ecte
d N
orm
al
for groupBullet= Group B - Experiment
Normal Q-Q Plot of Bullet PointReasoning
Most points are close to the line thus the assumption of normality is satisfied for the group A.
The graph shows curvature thus the assumption of normality is not satisfied for the group B.
Since for group B normality is not satisfied, cannot use t-test. Will go with nonparametric tests. Need to use Mann-Whitney test.
Descriptives
6.9524 .55410
6.8915
7.0000
6.448
2.53922
3.00
12.00
9.00
5.7917 .58353
4.5845
6.9988
5.4722
5.0000
8.172
2.85869
3.00
15.00
12.00
Mean
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Mean
5% Trimmed Mean
Median
Variance
Std. Deviation
Minimum
Maximum
Range
Groups for BulletPoint ReasoningGroup A - Control
Group B - Experiment
Bullet Point ReasoningStatistic Std. Error
Ranks
21 26.55 557.50
24 19.90 477.50
45
Groups for BulletPoint ReasoningGroup A - Control
Group B - Experiment
Total
Bullet Point ReasoningN Mean Rank Sum of Ranks
Test Statisticsa
177.500
477.500
-1.733
.083
Mann-Whitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
Bullet PointReasoning
Grouping Variable: Groups for Bullet Point Reasoninga.
According to above table, Mann-Whitney test value = -1.733, P-value = 0.083/2 = 0.0415 (need to divide the P-value by 2 as we have one-tailed test where as SPSS gives you 2-tailed test P-value.
Since (P-value = 0.0415) < ( = 0.05), null hypothesis is rejected.
Conclusion: At 5% level, Group B does have less bullet points than group A.
Tests of Normality
.258 21 .001 .873 21 .011
.276 24 .000 .741 24 .000
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 2Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Comparison Question #2 for Pre-test: On a scale from 1 to 5, where 1 represents not knowledgeable at all and 5 represents extremely knowledgeable, how knowledgeable are you with this topic?
Null: For Q2 answers do not differ significantly for group A and B.
Alternative: For Q2 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
Most points are close to the line thus the assumption of normality is satisfied for the group A.
4.03.53.02.52.01.51.0
Observed Value
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Pretest Q. 2
Group Statistics
21 2.5238 .98077 .21402
24 1.7500 .98907 .20189
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 2N Mean Std. Deviation
Std. ErrorMeanIndependent Samples Test
.223 .639 2.629 43 .012 .77381 .29439 .18012 1.36750
2.630 42.303 .012 .77381 .29422 .18017 1.36745
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 2F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)Mean
DifferenceStd. ErrorDifference Lower Upper
95% ConfidenceInterval of the
Difference
t-test for Equality of Means
54321
Observed Value
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
Exp
ecte
d N
orm
alfor groupPre= Group B - Experiment
Normal Q-Q Plot of Pretest Q. 2
Group B - ExperimentGroup A - Control
Groups for Pretest
5.00
4.00
3.00
2.00
1.00
Pre
test
Q. 2
42
Most points are close to the line except one thus the assumption of normality is satisfied for the group B.
Box plot shows outlier for group B. Even with the presence of outliers, normality is satisfied. Also this value is important for the analysis. Thus the decision is not to remove the outlier.
Independent Samples Test
.223 .639 2.629 43 .012
2.630 42.303 .012
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 2F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Tests of Normality
.218 21 .010 .904 21 .041
.277 24 .000 .810 24 .000
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 3Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.639) > ( = 0.05), thus assumption of equal variances is satisfied.
According to above table, t-test value = 2.629, P-value = 0.012.
Since (P-value = 0.012) < ( = 0.05), null hypothesis is rejected.
Conclusion: At 5% level, Q2 answers do differ significantly for group A and B.
Comparison Question #3 for Pre-test: On a scale from 1 to 5, where 1 represents not interested at all and 5 represents extremely interested, how interested are you in learning about analytic methodologies?
Null: For Q3 answers do not differ significantly for group A and B.
Alternative: For Q3 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
54321
Observed Value
1
0
-1
-2
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Pretest Q. 3
Most points are close to the line thus the assumption of normality is satisfied for the group A.
5.04.54.03.53.02.52.0
Observed Value
2.0
1.5
1.0
0.5
0.0
-0.5
-1.0
Exp
ecte
d N
orm
al
for groupPre= Group B - Experiment
Normal Q-Q Plot of Pretest Q. 3
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Group Statistics
21 3.4762 1.03049 .22487
24 2.8333 .91683 .18715
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 3N Mean Std. Deviation
Std. ErrorMeanIndependent Samples Test
.201 .656 2.215 43 .032
2.197 40.433 .034
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 3F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Group B - ExperimentGroup A - Control
Groups for Pretest
5.00
4.00
3.00
2.00
1.00
Pre
test
Q. 3
3
Box plot shows outlier for group A. Even with the presence of outliers, normality is satisfied. Also this value is important for the analysis. Thus the decision is not to remove the outlier.
Independent Samples Test
.201 .656 2.215 43 .032
2.197 40.433 .034
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 3F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Tests of Normality
.227 21 .006 .884 21 .018
.336 24 .000 .820 24 .001
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 1Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.656) > ( = 0.05), thus assumption of equal variances is satisfied.
According to above table, t-test value = 2.215, P-value = 0.032.
Since (P-value = 0.032) < ( = 0.05), null hypothesis is rejected.
Conclusion: At 5% level, Q3 answers do differ significantly for group A and B.
Comparison Question # 1 for Pre-test: On a scale from 1 to 5, where 1 represents not interested at all and 5 represents extremely interested, how interested are you in participating in this experiment?
Null: For Q1 answers do not differ significantly for group A and B.
Alternative: For Q1 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
5.04.54.03.53.02.52.0
Observed Value
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Pretest Q. 1
Most points are close to the line thus the assumption of normality is satisfied for the group A.
5.04.54.03.53.02.52.0
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
al
for groupPre= Group B - Experiment
Normal Q-Q Plot of Pretest Q. 1
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Group Statistics
21 3.4762 .98077 .21402
24 3.3333 .76139 .15542
Groups for PretestGroup A - Control
Group B - Experiment
Pretest Q. 1N Mean Std. Deviation
Std. ErrorMeanIndependent Samples Test
2.674 .109 .549 43 .586
.540 37.570 .592
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 1F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Independent Samples Test
2.674 .109 .549 43 .586
.540 37.570 .592
Equal variancesassumed
Equal variancesnot assumed
Pretest Q. 1F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Group B - ExperimentGroup A - Control
Groups for Pretest
5.00
4.50
4.00
3.50
3.00
2.50
2.00
Pre
test
Q. 1
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.109) > ( = 0.05), thus assumption of equal variances is satisfied.
Box plot does not show outliers for either group. This is a desired result for normality.
Tests of Normality
.277 21 .000 .860 21 .006
.297 24 .000 .830 24 .001
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 3Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
According to above table, t-test value = 0.549, P-value = 0.586.
Since (P-value = 0.586) > ( = 0.05), null hypothesis is not rejected.
Conclusion: At 5% level, Q1 answers do not differ significantly for group A and B.
Comparison Question #3 for Post-test: On a scale from 1 to 5, where 1 represents not at all and 5 represents extremely frequently, how frequently do you use structured analytic methods in solving intelligence related problems?
Null: For Q3 answers do not differ significantly for group A and B.
Alternative: For Q3 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
5.04.54.03.53.02.52.0
Observed Value
1.5
1.0
0.5
0.0
-0.5
-1.0
-1.5
Exp
ecte
d N
orm
alfor groupPre= Group A - Control
Normal Q-Q Plot of Posttest Q. 3
Most points are close to the line thus the assumption of normality is satisfied for the group A.
5.04.54.03.53.02.52.0
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
al
for groupPre= Group B - Experiment
Normal Q-Q Plot of Posttest Q. 3
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Group Statistics
21 3.1429 .91026 .19863
24 3.5000 .72232 .14744
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 3N Mean Std. Deviation
Std. ErrorMean
Independent Samples Test
.155 .696 -1.466 43 .150
-1.444 38.063 .157
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 3F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Group B - ExperimentGroup A - Control
Groups for Pretest
5.00
4.50
4.00
3.50
3.00
2.50
2.00
Po
stte
st Q
. 3Box plot does not show outliers for either group. This is a desired result for normality.
Independent Samples Test
.155 .696 -1.466 43 .150
-1.444 38.063 .157
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 3F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Tests of Normality
.241 20 .003 .862 20 .009
.299 24 .000 .812 24 .000
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 4Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.696) > ( = 0.05), thus assumption of equal variances is satisfied.
According to above table, t-test value = -1.466, P-value = 0.150.
Since (P-value = 0.150) > ( = 0.05), null hypothesis is not rejected.
Conclusion: At 5% level, Q3 answers do not differ significantly for group A and B.
Comparison Question #4 for Post-test: On a scale from 1 to 5, where 1 represents not at all and 5 represents a significant improvement, how much do you feel analytic methods improve intelligence analysis?
Null: For Q4 answers do not differ significantly for group A and B.
Alternative: For Q4 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
5.04.54.03.53.02.52.0
Observed Value
1.0
0.5
0.0
-0.5
-1.0
-1.5
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Posttest Q. 4
Most points are close to the line thus the assumption of normality is satisfied for the group A.
Group Statistics
20 3.9000 .96791 .21643
24 3.5417 .65801 .13431
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 4N Mean Std. Deviation
Std. ErrorMean
5.04.54.03.53.02.52.0
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
alfor groupPre= Group B - Experiment
Normal Q-Q Plot of Posttest Q. 4
Group B - ExperimentGroup A - Control
Groups for Pretest
5.00
4.50
4.00
3.50
3.00
2.50
2.00
Po
stte
st Q
. 4
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Box plot does not show outliers for either group. This is a desired result for normality.
Tests of Normality
.349 21 .000 .727 21 .000
.510 24 .000 .401 24 .000
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 6Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Independent Samples Test
1.345 .253 1.456 42 .153
1.407 32.474 .169
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 4F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.253) > ( = 0.05), thus assumption of equal variances is satisfied.
Independent Samples Test
1.345 .253 1.456 42 .153
1.407 32.474 .169
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 4F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
According to above table, t-test value = 1.456, P-value = 0.153.
Since (P-value = 0.153) > ( = 0.05), null hypothesis is not rejected.
Conclusion: At 5% level, Q4 answers do not differ significantly for group A and B.
Comparison Question #6 for Post-test: On a scale from 1 to 5, where 1 represents not enough time at all and 5 represents more than adequate amount of time, do you feel you had enough time to complete this process?
Null: For Q6 answers do not differ significantly for group A and B.
Alternative: For Q6 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
5.04.54.03.53.0
Observed Value
0.6
0.3
0.0
-0.3
-0.6
-0.9
-1.2
-1.5
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Posttest Q. 6
Most points are close to the line thus the assumption of normality is satisfied for the group A.
5.04.54.03.53.0
Observed Value
0.0
-0.5
-1.0
-1.5
Exp
ecte
d N
orm
al
for groupPre= Group B - Experiment
Normal Q-Q Plot of Posttest Q. 6
Descriptives
4.4286 .16288
.557
.74642
3.00
5.00
4.8333 .09829
.232
.48154
3.00
5.00
Mean
Variance
Std. Deviation
Minimum
Maximum
Mean
Variance
Std. Deviation
Minimum
Maximum
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 6Statistic Std. Error
Most points are not close to the line thus the assumption of normality is not satisfied for the group B.
We cannot use the t-test. Need to use nonparametric test, Mann-Whitney test.
Ranks
21 19.36 406.50
24 26.19 628.50
45
Groups for PretestGroup A - Control
Group B - Experiment
Total
Posttest Q. 6N Mean Rank Sum of Ranks
Tests of Normality
.223 20 .010 .809 20 .001
.261 24 .000 .872 24 .006
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 7Statistic df Sig. Statistic df Sig.
Kolmogorov-Smirnova
Shapiro-Wilk
Lilliefors Significance Correctiona.
Test Statisticsa
175.500
406.500
-2.248
.025
Mann-Whitney U
Wilcoxon W
Z
Asymp. Sig. (2-tailed)
Posttest Q. 6
Grouping Variable: Groups for Pretesta.
According to above table, Mann-Whitney test value = 175.5, P-value = 0.025.
Since (P-value = 0.025) < ( = 0.05), null hypothesis is rejected.
Conclusion: At 5% level, Q6 answers do differ significantly for group A and B.
Comparison Question #7 for Post-test: On a scale from 1 to 5, where 1 represents no enjoyment at all and 5 represents thorough enjoyment, how much did you enjoy this process?
Null: For Q7 answers do not differ significantly for group A and B.
Alternative: For Q7 answers do differ significantly for group A and B. – claim
Will be using t-test for independent samples.
Testing normality assumption as sample sizes are < than 30.
Group Statistics
20 3.9000 .78807 .17622
24 3.4583 .93153 .19015
Groups for PretestGroup A - Control
Group B - Experiment
Posttest Q. 7N Mean Std. Deviation
Std. ErrorMean
Independent Samples Test
.685 .412 1.678 42 .101
1.704 41.984 .096
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 7F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Both Kolmogorov-Smirnov and Shapiro-Wilk tests give p-value < ( = 0.05), thus normality assumption is not satisfied for both Groups. Need to check Normal Probability plots for both groups.
5.04.54.03.53.0
Observed Value
1.0
0.5
0.0
-0.5
-1.0
Exp
ecte
d N
orm
al
for groupPre= Group A - Control
Normal Q-Q Plot of Posttest Q. 7
54321
Observed Value
2
1
0
-1
-2
Exp
ecte
d N
orm
al
for groupPre= Group B - Experiment
Normal Q-Q Plot of Posttest Q. 7
Most points are close to the line thus the assumption of normality is satisfied for the group A.
Most points are close to the line thus the assumption of normality is satisfied for the group B.
Independent Samples Test
.685 .412 1.678 42 .101
1.704 41.984 .096
Equal variancesassumed
Equal variancesnot assumed
Posttest Q. 7F Sig.
Levene's Test forEquality of Variances
t df Sig. (2-tailed)
t-test for Equality of Means
Here need to check if the assumption of equal variances is satisfied.
According to Levene’s test (P-value = 0.412) > ( = 0.05), thus assumption of equal variances is satisfied.
According to above table, t-test value = 1.678, P-value = 0.101.
Since (P-value = 0.101) > ( = 0.05), null hypothesis is not rejected.
Conclusion: At 5% level, Q7 answers do not differ significantly for group A and B.