# Decision Analysis

Y. İlker TOPCU, Ph.D.

www.ilkertopcu.info www.yoneylem.itu.edu.tr

Outline

Introduction

**What is Decision Analysis?
**

Decision Making under Certainty

**Decision Making under Uncertainty
**

Decision Making under Risk Utility Theory Decision Trees

Introduction

• One dimensional (single criterion) decision making • Single stage vs. multi stage decision making

• Decision analysis is an analytical and systematic way to tackle problems • A good decision is based on logic (rational decision maker)

**Components of Decision Analysis
**

• A state of nature is an actual event that may occur in the future. • A payoff matrix (decision table) is a means of organizing a decision situation, presenting the payoffs from different decisions/alternatives given the various states of nature.

**Basic Steps in Decision Analysis
**

1) Clearly define the problem at hand 2) List the possible alternatives 3) Identify the possible state of natures (outcomes) 4) List the payoff or profit/cost of alternatives with respect to state of natures 5) Select one of the mathematical decision analysis models

6) Apply the model and make your decision

**Types of Decision-Making Environments
**

• Type 1: Decision-making under certainty • The decision-maker knows with certainty the consequences of every alternative or decision choice • Type 2: Decision-making under uncertainty

**• The decision-maker does not know the probabilities of
**

the various outcomes. Actually s/he knows nothing! • Type 3: Decision-making under risk • The decision-maker does know the probabilities of the various outcomes

**Decision Making Under Certainty
**

• Instead of state of natures, a true state is known to the decision maker before s/he has to make decision

• The optimal choice is to pick an alternative with the highest payoff

**Decision Making Under Uncertainty
**

• Maximax • Maximin • Criterion of Realism • Equally likelihood

• Minimax

Payoff Matrix

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $200,000 ($180,000) Construct small plant $100,000 ($20,000) Do nothing $0 $0

Maximax

Choose the alternative with the maximum optimistic level

m m

m

**ok = max {oi} = max{ max{vij}} i 1 i 1 j 1
**

STATES OF NATURE Favorable Unfavorable ALTERNATIVES Construct large plant Construct small plant Do nothing market $200,000 $100,000 $0 market ($180,000) ($20,000) $0 Maximum in row (ok) $200,000 $100,000 $0

Maximin

Choose the alternative with the maximum security level

m m

**sk = max {si} = max { min {vij}} i 1 i 1 j 1
**

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $200,000 ($180,000) Construct small plant $100,000 ($20,000) Do nothing $0 $0 Minimum in row (sk ) ($180,000) ($20,000) $0

m

Criterion of Realism

Hurwicz suggested to use the optimism-pessimism index (a) Choose the alternative with the maximum weighted average of optimistic and security levels

max {a oi + (1 – a) si} where 0<a<1

i 1 m

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $200,000 ($180,000) Construct small plant $100,000 ($20,000) Do nothing $0 $0

Weighted average a=0.7 380a-180 86 120a-20 64 $0 0

Criterion of Realism

120a-20 = 0 a = 0.1667 380a-180 = 120a-20 a = 0.6154

0<a<0.1667 “Do nothing” 0.1667<a<0.6154 “Construct small plant” 0.6154<a<1 “Construct large plant”

Equally Likelihood

Laplace argued that “knowing nothing at all about the true state of nature” is equivalent to “all states having equal probability” Choose the alternative with the maximum row average (expected value)

STATES OF NATURE Favorable Unfavorable market market $200,000 ($180,000) $100,000 ($20,000) $0 $0 Row average $10,000 $40,000 $0

ALTERNATIVES Construct large plant Construct small plant Do nothing

Minimax

Savage defined the regret (opportunity loss) as the difference between • the value resulting from the best action given that state of nature j is the true state and • the value resulting from alternative i with respect to state of nature j

**Choose the alternative with the minimum worst (maximum) regret
**

Regret Values STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $0 $180,000 Construct small plant $100,000 $20,000 Do nothing $200,000 $0

Maximum in row $180,000 $100,000 $200,000

**Summary of Example Results
**

CRITERION • Maximax • Maximin • Criterion of Realism • Equally likelihood • Minimax DECISION “Construct large plant” “Do nothing” depends on a “Construct small plant” “Construct small plant”

The appropriate criterion is dependent on the personality and philosophy of the decision maker.

Solutions with QM for Windows

**Decision Making Under Risk
**

Probability

• Objective (frequentist approach) • Subjective

**• Expected (Monetary) Value
**

Expected Value of Perfect Information

**• Expected Opportunity Loss • Utility Theory
**

Certainty Equivalence Risk Premium

Probability

• Life is uncertain! • We must deal with risk! • A probability is a numerical statement about the likelihood that an event will occur

**Basic Statements About Probability
**

• The probability, P, of any event or state of nature occurring is greater than or equal to 0 and less than or equal to 1. That is: 0 P(event) 1 • The sum of the simple probabilities for all possible outcomes of an activity must equal 1

Objective Probability

Determined by experiment or observation:

• Probability of heads on coin flip • Probability of spades on drawing card from deck

P ( event ) Number of times event occurs Total number of outcomes or occurrences

P(A): probability of occurrence of event A n(A): number of times that event A occurs n: number of independent and identical repetitions of the experiments or observations

P(A) = limit n(A) / n n

Subjective Probability

• Determined by an estimate based on

• • • • personal belief, judgment experience, knowledge of a situation

• Probability is taken as representing the decision maker’s degree (or strength) of belief that the system will adopt a certain state.

Payoff Matrix with Probabilities

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Probabilites 60% 40% Construct large plant $200,000 ($180,000) Construct small plant $100,000 ($20,000) Do nothing $0 $0

**Expected Value Criterion
**

Choose the alternative with the maximum weighted row average EV(ai) = vij P(qj)

j

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $200,000 ($180,000) Construct small plant $100,000 ($20,000) Do nothing $0 $0 PROBABILITIES 0.6 0.4

Expected Value $48,000 $52,000 $0

Sensitivity Analysis

EV(Large Plant) = $200,000P - $180,000 (1-P) EV(Small Plant) = $100,000P - $20,000(1-P) EV(Do Nothing) = $0P + $0(1-P)

Sensitivity Analysis

250000 200000 150000 100000 50000 0 -50000 -100000 -150000 -200000

Point 1 Small Plant

Point 2

EV Values

0

0.2

0.4

0.6

0.8

1

Large Plant EMV Values of P

**Expected Value of Perfect Information
**

• A consultant or a further analysis can aid the decision maker by giving exact (perfect) information about the true state: the decision problem is no longer under risk; it will be under certainty. • Is it worthwhile for obtaining perfectly reliable information: is EVPI greater than the fee of the consultant (the cost of the analysis)? • EVPI is the maximum amount a decision maker would pay for additional information

**Expected Value of Perfect Information
**

EVPI = Expected payoff with perfect information – Best expected payoff without perfect information

ALTERNATIVES Construct large plant Construct small plant Do nothing PROBABILITIES STATES OF NATURE Favorable Unfavorable market market $200,000 ($180,000) $100,000 ($20,000) $0 $0 0.6 0.4 Expected Value $48,000 $52,000 $0

In the example: Expected payoff with perfect information: 200*.6+0*.4=120 Best expected value: 52 EVPI = 120 – 52 = 68

**Expected Opportunity Loss Criterion
**

Choose the alternative with the minimum weighted row average of the regret matrix EOL(ai) = rij P(qj)

j

Regret Values

STATES OF NATURE Favorable Unfavorable ALTERNATIVES market market Construct large plant $0 $180,000 Construct small plant $100,000 $20,000 Do nothing $200,000 $0 PROBABILITIES 0.6 0.4

Expected Opportunity Loss $72,000 $68,000 $120,000

Utility Theory

• Utility assessment assigns the worst outcome a utility of 0, and the best outcome, a utility of 1. • A standard gamble is used to determine utility values: When you are indifferent, the utility values are equal. • Choose the alternative with the maximum expected utility EU(ai) = u(ai) = u(vij) P(qj)

j

**Standard Gamble for Utility Assessment
**

(p) Best outcome (v*) u(v*) = 1

(1–p)

Worst outcome (v–) u(v–) = 0 Certain outcome (v) u(v) = 1*p+0*(1–p)

**Utility Assessment (1st approach)
**

I

(0.5) Best outcome (v*) u(v*) = 1 Worst outcome (v–) u(v–) = 0 Certain outcome (x1) u(x1) = 0.5 (0.5)

II

(0.5)

v* u(v*) = 1

(0.5)

(0.5)

x1 u(x1) = 0.5 x2 u(x2) = 0.75

III

(0.5)

x1 u(v*) = 0.5 Worst outcome (v–) u(v–) = 0 x3 u(x3) = 0.25

**In the example: u(-180) = 0 and u(200) = 1 X1= 100 u(100) = 0.5 X2 = 175 u(175) = 0.75 X3 = 5 u(5) = 0.25
**

1 0.8 0.6 0.4 0.2 0 -200 -150 -100 -50 0 50 100 150 200

**Utility Assessment (2nd approach)
**

(p) Best outcome (v*) u(v*) = 1 Worst outcome (v–) u(v–) = 0 Certain outcome (vij) u(vij) = p

(1–p)

In the example: u(-180) = 0 and u(200) = 1 For vij=–20, p=%70 u(–20) = 0.7 For vij=0, p=%75 u(0) = 0.75 For vij=100, p=%90 u(100) = 0.9

1 0.8 0.6 0.4 0.2 0 -200 -150 -100 -50 0 50 100 150 200

Utilities ALTERNATIVES Construct large plant Construct small plant Do nothing PROBABILITIES

STATES OF NATURE Favorable Unfavorable market market 1 0 0.9 0.7 0.75 0.75 0.6 0.4

Expected Utility 0.6 0.82 0.75

Preferences for Risk

Utility

Monetary Outcome

Certainty Equivalence

• If a DM is indifferent between accepting a lottery ticket and accepting a sum of certain money, the monetary sum is the Certainty Equivalent (CE) of the lottery

• Z is CE of the lottery (Y>Z>X).

p X

1–p Y

Z

Risk Premium

• Risk Premium (RP) of a lottery is the difference between the EV of the lottery and the CE of the lottery

• If the DM is risk averse, RP>0

S/he prefers to receive a sum of money equal to expected value of a lottery than to enter the lottery itself

**• If the DM is risk prone, RP<0
**

S/he prefers to enter a lottery than to receive a sum of money equal to its expected value

**• If the DM is risk neutral, RP=0
**

S/he is indifferent between entering any lottery and receiving a sum of money equal to its expected value