Professional Documents
Culture Documents
Techniques
Decision Making under
Risk & Uncertainty
Decision Environments
Decision Environments
• Certainty
• Risk
• Uncertainty
Dr. M S Mahapatra
Decision Making Under Certainty
Decision Models:
• Linear Programming
• Analytic Hierarchy Process (AHP)
• EOQ and EPQ Models
Dr. M S Mahapatra
Decision Making Under Risk
Alternatives and Outcomes
Known
Alternative Outcomes
Known only with a probability associated with them. However, the
probabilities are known.
Dr. M S Mahapatra
Decision Making Under Uncertainty
Alternative Outcomes
Not known – even the probabilities associated with them are
not known.
Dr. M S Mahapatra
Criteria for Decision Making under
Uncertainty
1. Maximax or Optimism Criterion
Dr. M S Mahapatra
Maximax: Optimism Criterion
• Find the maximum possible payoff for each decision
alternative
• Select the decision alternative with the maximum of the
above maximum payoffs
STATE OF NATURE
No Favorable Unfavorable Payoff for a
Decision Market market Market Decision
Alternatives Changes changes changes Strategy
Redesign on a
30 100 –80
small scale
Rebuild and
50 200 –200
Refurbish
Do nothing 0 30 –50
Dr. M S Mahapatra
Maximin: Pessimism Criterion
• Find the minimum possible payoff for each decision
alternative
• Select the decision alternative with the maximum of the
above minimum payoffs
STATE OF NATURE
No Favorable Unfavorable Payoff for a
Decision Market market Market Decision
Alternatives Changes changes changes Strategy
Redesign on a
30 100 –80
small scale
Rebuild and
50 200 –200
Refurbish
Do nothing 0 30 –50
Dr. M S Mahapatra
Hurwicz: Criterion of Realism
• Choose a degree of Optimism (a). Obtain Maximum Payoff
(Max) and Minimum Payoff (Min) for each decision alternative.
Obtain Hurwicz Payoffs for each strategy as
a*Max + (1 – a)*Min
• Select the decision alternative with the maximum of the above
Hurwicz payoffs
STATE OF NATURE
No Favorable Unfavorable Payoff for a
Decision Market market Market Decision
Alternatives Changes changes changes Strategy
Redesign on a
30 100 –80
small scale
Rebuild and
50 200 –200
Refurbish
Do nothing 0 30 –50
Dr. M S Mahapatra
Laplace: Equally Likely Criterion
• Find the average value of payoff for each decision alternative
• Select the decision alternative with the maximum of the average
payoff values
STATE OF NATURE
No Favorable Unfavorable Payoff for a
Decision Market market Market Decision
Alternatives Changes changes changes Strategy
Redesign on a
30 100 –80
small scale
Rebuild and
50 200 –200
Refurbish
Do nothing 0 30 –50
Dr. M S Mahapatra
Savage: Minimax Regret Criterion
• Find the regret matrix (For each payoff under a state of nature,
compute the difference from the maximum payoff). Obtain
maximum regret for each decision alternative
• Select the decision alternative with the minimum of the above
maximum regrets
STATE OF NATURE
No Favorable Unfavorable Payoff for a
Decision Market market Market Decision
Alternatives Changes changes changes Strategy
Redesign on a
30 100 –80
small scale
Rebuild and
50 200 –200
Refurbish
Do nothing 0 30 –50
Dr. M S Mahapatra
Criteria for Decision Making under
Risk
Steps:
1) Obtain Probability of Occurrence (𝑝𝑝𝑖𝑖 ) for each state of
nature (i).
2) Find the Expected Value for each decision alternative as
the sum of (𝑝𝑝𝑖𝑖 x 𝑝𝑝𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑖𝑖 ) for all the states of nature.
3) Select the decision alternative with the best expected value.
Dr. M S Mahapatra
Expected Value Approach
STATE OF NATURE
No Favorable Unfavorable Expected
Decision Market market Market Value for a
Alternatives Changes changes changes Decision
(0.3) (0.4) (0.3) Strategy
Redesign on a
30 100 –80 25
small scale
Rebuild and
50 200 –200 35
Refurbish
Do nothing 0 30 –50 -3
Dr. M S Mahapatra
Sensitivity Analysis
A payoff matrix is given below with three decision alternatives and two states
of nature. At present, the decision strategy “Constructing a small plant” shows
the largest expected value.
Perform the sensitivity analysis for this decision problem.
STATE OF NATURE
FAVORABLE UNFAVORABLE
ALTERNATIVE MARKET ($) MARKET ($) EV ($)
Dr. M S Mahapatra
Sensitivity Analysis
Dr. M S Mahapatra
Sensitivity Analysis
$300,000
Point 2
0 EV (do nothing)
0.167 0.615 1
–$100,000 Probability (P)
–$200,000
Dr. M S Mahapatra
Sensitivity Analysis
Point 1:
EV(do nothing) = EV(small plant)
20,000
0 = $120,000 P − $20,000 P= = 0.167
120,000
Point 2:
EV(small plant) = EV(large plant)
$120,000 P − $20,000 = $380,000 P − $180,000
160,000
P= = 0.615
260,000
BEST RANGE OF P
ALTERNATIVE VALUES
Do nothing Less than 0.167
Dr. M S Mahapatra
Expected Value of Perfect Information
(EVPI)
Dr. M S Mahapatra
Expected Value of Perfect Information
(EVPI)
where,
EREV: Expected Return of the EV approach
ERPI: Expected Return with Perfect Information
ERPI =
(best payoff for the first state of nature) x (probability of the first state of nature) +
(best payoff for the second state of nature) x (probability of the second state of
nature) + … + (best payoff for the last state of nature)x (probability of the last
state of nature
Dr. M S Mahapatra
Expected Value of Perfect Information
State of Nature
Alternative Favorable Unfavorable Expected Value (EV)
Market ($) Market ($)
Construct a large
200,000 -180,000 10,000
plant
Construct a small
100,000 -20,000 40,000
plant
Do nothing 0 0 0
Probability 0.5 0.5
Perfect
200,000 0 ERPI = 100,000
Information
Compute ERPI
State of Nature: Favorable Market, Best Payoff: 200,000
State of Nature: Unfavorable Market, Best Payoff: 0
ERPI = (200,000)(0.5) + (0)(0.5) = $100,000
Compute EVPI
EVPI = ERPI – max EREV = $100,000 − $40,000 = $60,000 - the most we should
pay for additional information.
Dr. M S Mahapatra
Decision Trees
A Decision Tree represents decisions and outcomes in sequential
(chronological) order
Symbols
A decision node from which one of several alternatives may be
selected.
A state of nature node (chance node) out of which one state of
nature will occur.
An arc (branch) represents possible decisions or states of nature
Dr. M S Mahapatra
Example
STATE OF NATURE
FAVORABLE UNFAVORABLE
ALTERNATIVE MARKET ($) MARKET ($)
Construct a large plant 200,000 –180,000
Do nothing 0 0
Dr. M S Mahapatra
Decision Tree: Example
Probability Payoff
A State-of-Nature Node
Favorable Market (0.5)
$200,000
A Decision Node
1
Unfavorable Market (0.5)
–$180,000
$0
Dr. M S Mahapatra
Decision Tree: Example
Probability Payoff
EV for Node 1= $10,000
Favorable Market (0.5)
$200,000
1
Unfavorable Market (0.5)
–$180,000
$0
Dr. M S Mahapatra
Decision Making with Experimentation
STATE OF NATURE
FAVORABLE UNFAVORABLE
ALTERNATIVE MARKET ($) MARKET ($)
Construct a large plant 200,000 –180,000
Construct a small plant 100,000 –20,000
Do nothing 0 0
Probability 0.5 0.5
• Before deciding about building a new plant, the company has the option to
conduct its own marketing survey, at a cost of $10,000.
• Information from the survey could help in deciding which alternative to
pursue (large, small, or no plant)
Dr. M S Mahapatra
The Experimentation
• The market survey is the experimentation conducted at a cost of $10,000.
• However, the market survey will not be able to predict the state of nature with 100%
accuracy.
• The market survey can only give additional information on the probable states of nature
in the form of the following conditional probabilities:
Dr. M S Mahapatra
Bayesian Analysis for Probability
Survey Fav P=0.7
Fav. Market
P(FM)=0.50 Survey Unfav P=0.3
Dr. M S Mahapatra
Bayesian Analysis for Probability
Survey Fav P=0.7
P(FM∩SF)=0.35
Fav. Market
P(FM)=0.50 Survey Unfav P=0.3
P(FM∩SU)=0.15
P(UM∩SF)=0.10
Unfav. Market Survey Fav P=0.2
P(UM)=0.50 P(UM∩SU)=0.40
Survey Unfav P=0.8
Joint Probabilities:
P(FM∩ SF) = 0.5x0.7 = 0.35
Thus, Unconditional Probabilities:
P(FM ∩ SU) = 0.5x0.3 = 0.15
P(Survey Favorable) = 0.35 + 0.10 = 0.45
P(UM∩ SF) = 0.5x0.2 = 0.10
P(Survey Unfavorable) = 0.15+0.40 = 0.55
P(UM∩ SU) = 0.5x0.8 = 0.40
Dr. M S Mahapatra
Bayesian Analysis for Probability
Survey Fav P=0.7
P(FM∩SF)=0.35
Fav. Market
P(FM)=0.50 Survey Unfav P=0.3
P(FM∩SU)=0.15
P(UM∩SF)=0.10
Unfav. Market Survey Fav P=0.2
P(UM)=0.50 P(UM∩SU)=0.40
Unconditional Probabilities: Survey Unfav P=0.8
P(Survey Favorable) = 0.35 + 0.10 = 0.45 P(Survey Unfavorable) = 0.15+0.40 = 0.55
So, Posterior Probabilities:
P(Fav. Market | Survey Favorable) = P(FM∩SF) / P(SF) = 0.35/0.45 = 0.78
P(Unfav. Market | Survey Favorable) = P(UM ∩SF)/ P(SF) = 0.10/0.45 = 0.22
P(Fav. Market | Survey Unfavorable) = P(FM ∩SU)/ P(SU) = 0.15/0.55 = 0.27
P(Unfav. Market | Survey Unfavorable) = P(UM ∩SU)/ P(SU) = 0.40/0.55 = 0.73
Dr. M S Mahapatra
Decision Tree of the Problem
Note: the $10,000 cost was subtracted from each of the first 10 branches. The,
$190,000 payoff was originally $200,000 and the $-10,000 was originally $0.
First Second Decision Payoffs
Decision Point
Point
Favorable Market (0.78)
$190,000
2 Unfavorable Market (0.22)
–$190,000
Favorable Market (0.78)
Small $90,000
Plant
3 Unfavorable Market (0.22)
–$30,000
No Plant
–$10,000
1 Favorable Market (0.27)
$190,000
4 Unfavorable Market (0.73)
–$190,000
Favorable Market (0.27)
Small $90,000
Plant
5 Unfavorable Market (0.73)
–$30,000
No Plant
–$10,000
Dr. M S Mahapatra
Estimation of the payoffs
1. Given favorable survey results (positive survey)
EV(node 2) = EV(large plant | positive survey)
= (0.78)($190,000) + (0.22)(–$190,000) = $106,400
EV(node 3) = EV(small plant | positive survey)
= (0.78)($90,000) + (0.22)(–$30,000) = $63,600
EV for no plant = –$10,000
If survey results are favorable, a large plant should be built with an expected value of $106,400
Dr. M S Mahapatra
32
Estimation of the Payoffs
3. Thus, the expected value of the market survey,
EV(node 1) = EV(conduct survey)
= (0.45)($106,400) + (0.55)($2,400) = $49,200
If the market survey is not conducted, a small plant should be built with an expected
value of $40,000
$106,400
Small 3 $90,000
Unfavorable Market (0.22)
Plant –$30,000
No Plant –$10,000
$49,200
1 –$87,400 Favorable Market (0.27) $190,000
4 Unfavorable Market (0.73) –$190,000
$2,400 Favorable Market (0.27)
$2,400 Small 5 $90,000
Unfavorable Market (0.73) –$30,000
Plant
No Plant –$10,000
$49,200
Small 7 $100,000
Unfavorable Market (0.50)
Plant –$20,000
No Plant
$0
Dr. M S Mahapatra