EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
1. Decision Analysis
Why are decisions hard?
1.
2.
3.
4.
5.
Complexity
Uncertainty of Key Elements
Multiple Objectives
Different Perspectives
Sensitivity/Unstability
Why study decision analysis?
1.
2.
3.
4.
5.
Supplies methods for organizing decisions
Allows identification of important sources of uncertainty
Forces representation of uncertainty
Supplies framework for dealing with multiple objectives
Modeling & Sensitivity Analysis allows to sort through
the problem
Main Objective
Decision Analysis leads to "better" decisions
Better:
1. Decisions are consistent
2. No surprises due to through study of the problem
3. Performance of decision making is better on the average
A good decision = Looking back in the past,
one can say that one would have made the
same decision given the information available
at the time of the decision.
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 1
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Definition Decision Analysis
Prescriptive approach for people who want to think
hard and systematically about decision problems.
Comments on decision analysis:
1. A decision analysis (DA) is an information source
2. A DA should not replace a decision maker but
should support him
3. A DA does not only provide solution, but also
provides insight to:




Situation
Uncertainty
Objectives
Trade off
A Decision Analysis can only yield
a recommended course of action
Important inputs for decision analysis:
1.
2.
Subjective Judgement about uncertainty
Subjective Judgement about preferences
BE CAREFULL, human beings are
imperfect information processors
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 2
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Flow Chart of Decision Analysis
Identify the problem
Identify:
Address problem
or keep people happy
Identify objectives
and alternatives
Objectives:
Min. cost? Max. profit?
Min. risk?
Max. market share?
Decompose and model
the problem:
• Model of problem structure
• Model of uncertainty
• Model of preferences
Decompose:
“Divide & Conquer”
Mathematical Models
are helpful.
Choose the
best alternative
Sensitivity:
“What if?”
Does optimal decision
change?
Sensitivity Analysis
Is further
analysis
needed?
Alternatives:
Invest, not invest?
Partially invest?
Attack, not attack?
Partially attack
YES
NO
Further Analysis:
New objectives?
New Alternatives?
Changed insight in
uncertainties?
Changed insight in
preference?
Recommendation: Implement
the chosen alternative
Source: Making Hard Decisions by Clemen
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 3
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Output Decision Analysis process:
Requisite Decision Model
“A model is requisite if no new intuitions emerge about the
problem or when it contains everything essential for solving
the problem”
Convergence to requisite decision model from:
 Technical modeling expertise of the analyst
 The will of the decision maker not to accept incomplete or
inappropriate models
Where is Decision Analysis Used?:
 Business
 Government
Examples:
1. Managing Research and Development Programs.
2. Understanding the World Oil Market.
3. Forecasting Sales for a new Product.
4. Electric Power Generation.
5. Deciding whether to launch a new product or venture
 Medicine
1. Help doctors make specific diagnosis
2. Optimal Inventory of blood in a blood bank
3. Firms decision regarding different kinds of medical
insurance.
Corner and Kirkwood (1991) - "Decision Analysis
Applications in the Operations Research Literature".
Operations Research, Vol. 39, 206 - 219
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 4
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Future Subjects
1. Modeling Decisions:
 Development of Influence Diagrams
 Development of Decision Trees
2. Modeling Uncertainty – A
 Review Basic Probability
 Solving Decision Trees
 Sensitivity Analysis
 Subjective Probability
3. Midterm Exam
4. Modeling Uncertainty – B




Theoretical Probability Models
Using data in Decision Problems
Combining Data & Subjective Probability
Monte Carlo Analysis of Decision Problems
5. The Value of Information
 Value of Perfect Information
 Value of Imperfect Information
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 5
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
2. Elements of Decision Problems
1. Objectives
2. Decisions
3. Uncertain events
4. Consequences
1. Objectives:
A specific thing one wants to achieve:
 To harvest successfully – Farmer.
 To resolve a specific scientific problem – Scientist.
 To make a lot of money – Investor.
Person’s Values: Collection of objectives
No values
No objectives
No Decisions
Special Objective: Making Money; often a surrogate
objective (or means objective) for other objectives e.g.
eating, traveling, afford clothing, etc.
Underlying Objectives: Understanding the underlying
objectives (or fundamental objectives) is crucial to
formulating the decision problem and list alternatives.
Often a trade off has to be made between “money making”
objective and other fundamental objectives.
Decision Context and Values:
DECISION CONTEXT =
The setting in which a decision occurs
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 6
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Person Value’s = Collection of All Objectives
Decision Context activates only subset of objectives.
Super Computer
Objectives
Price
•Five-Year Costs
•Cost of Improved
performance
Performance
•Speed
•Throughput
•Memory Size
•Disk Size
•On-Site performance
User Needs
•Installation Date
•Rollin/Roll out
•Ease of Use
•Software
Compatibility
•Mean Time to
Failure
Operational Needs
•Square Footage
•Water Cooling
•Operator Tools
•Telecommunications
•Vendor Support
Management Issues
•Vendor Health
•US Ownership
•Commitment to
super computer
Source: Making Hard Decisions by Clemen
2. Decisions
Most decisions call for: The Immediate Decision
 Go, No Go
 Accept bet or not
 Invest amount of money – may vary in range of values
Look at all possible decision alternatives: Be Creative!
 Do nothing
 Wait for information and decide later on
 Execute Hedging
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 7
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Decision may be of a sequential nature:
Future decisions may need to be taken into
account when making immediate decision
Future decisions may depend on:
 Past Decisions
 Past Events
 Both
List of possible decisions is important, but more important
is the order in which they occur.
Immediate
Decision
Second
Decision
Third
Decision
Final
Decision
Results
are
known
Now
Time Line
Consecutive decisions can be the same, but may differ.
3. Uncertain Events
Many decisions (All?) are made
under the presence of uncertainty
 Investment decision: Will stock of company go up or not?
 Camping decision: Will the weather be good or not?
 Mutual fund decision: Will entire stock market go up?
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 8
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
A decisions problem becomes more complicated when
the number of relevant uncertain events increases.
Nature of Uncertain Events:
 Outcomes are measured in distinct classes (Discrete)
 Outcomes may fall in a range of values (Continuous)
 Interdependence between different uncertain events
The time sequence of uncertain events related
to the sequence of decision is important.
Why?
 Tells you what information becomes known before a
decision has to be made
 Uncertain events may be unknown at the time of the
immediate decision, but may be known by subsequent
decisions
Resolved before
second decision
Immediate
Decision
Resolved before
third decision
Second
Decision
Resolved before
final decision
Third
Decision
Resolved before
last decision
Final
Decision
Results
are
known
Now
Time Line
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 9
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
4. Consequences
Evaluation of outcomes:
 Profit – Measured in # Dollars
 Casualties – Measured in # Deaths
 Environmental Damage – Measured in # Polluted Soil
 Health Risk – Measured in # Infected People
Trade off between has to be made
in almost any decision problem
 Profit, Casualties, Environmental Damage, Health Risk –
Measured in # Utils.
Planning Horizon = Time when decision maker
finds out the results
Resolved before
second decision
Resolved before
third decision
Resolved before
final decision
Resolved before
last decision
Consequences
Immediate
Decision
Second
Decision
Third
Decision
Final
Decision
Planning
Horizon
Now
Time Line
For requisite decision models:
Stop when future decisions and future uncertain events are
not essential to the immediate decision.
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 10
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Time Value of Money: A Special kind of Trade-Off
$100 @ 10% annual interest now =$110 one year later
$110 @ 10% annual interest now =$121 one year later
or
$100 @ 10% annual interest now =$121 two years later
General Formulation:
PV(X)
= Present Value of dollar amount X
R%
= interest per period
FVn(X)
= Future Value of X after n periods
R n
FVn ( X )  (1 
)  PV ( X )
100
At immediate decision:
Future Values must be converted to Present Values
PV ( X ) 
FVn ( X )
R n
(1 
)
100
Note difference between:
1. 10% Annual Interest Rate
2. 10% Annual Interest Rate compounded monthly
1. $100 @10% Annual Interest = $110 dollars after one year
2. $100 @ 10/12% Monthly Interest over 12 periods
12
10 

 * $100 = $110.47
= 1 
 12 *100 
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 11
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen
EMSE 269 - Elements Problem Solving and Decision Making
7/29/17
Stream of Cash Flows:
$425 in savings @ 10% annual interest
Pay $425 and receive $110, $121, $133.10
$146.41 after 1st, 2nd, 3rd, 4th year respectively
What is the preferred alternative?
Year
1
Future Value
$110
2
$121
3
$133.10
4
$146.41
Total
$510.41
Present Value
110
 100
1.1
$121
 $100
(1.1) 2
$133.10
 $100
(1.1) 3
$146.41
 $100
(1.1) 4
$400
General Formulation:
Xk = cost/revenue at the end of period k
R% = interest rate per period
n
Xk
NPV ( X 0 , X 1 , X n )  
k
R 
k 0 
1 

 100 
 Note: cost
Xk negative ;revenue
Xk positive
Instructor: Dr. J. Rene van Dorp
Session 1 - Page 12
Source: Making Hard Decisions, An Introduction to Decision Analysis by R.T. Clemen