Unit 5
Decision
Analysis
Learning Objectives
• Learn about decision making under certainty, under uncertainty, and
under risk.
• Learn several strategies for decision-making under uncertainty,
including expected payoff, expected opportunity loss, maximin,
maximax, and minimax regret.
• Learn how to construct and analyze decision trees.
• Understand aspects of utility theory.
• Learn how to revise probabilities with sample information.
Decision-Making Scenarios
• Decision-making under certainty
• Decision-making under uncertainty
• Decision-making under risk
Three Variables
in Decision Analysis Problems
• Decision Alternatives are the various choices or
options available to the decision maker in any
given problem situation
• States of Nature are the occurrences of nature
that can effect the outcome of the decision. They
are beyond the decision maker’s control
• Payoffs are the benefits or rewards that result
from selecting a particular decision alternative.
They are often expressed in dollars, but may be
stated in other units, such as market share
Decision Table
States of Nature
s1 s2 s3 s n
d 1 P1,1 P1,2 P1,3 P 1, n
Decision d 2 P2,1 P2,2 P2,3 P 2, n
Alternatives
d 3 P3,1 P3,2 P3,3 P 3, n
d m Pm,1 Pm,2 Pm,3 P
m, n
where: sj = state of nature
dj = decision alternative
Pi,j = payoff for decision i under state j
Example: Decision Table
for an Investor
Slow Rapid
Stagnant Growth Growth
Stocks $ (500) $ 700 $ 2,200
Bonds $ (100) $ 600 $ 900
CDs $ 300 $ 500 $ 750
Mixture $ (200) $ 650 $ 1,300
Annual payoffs for an investment of $10,000
Decision-Making Under Certainty
The states of nature are known. The
Theeconomy
economy
will
willgrow
grow
rapidly.
rapidly.
Slow Rapid Invest
Investin
instocks.
stocks.
Stagnant Growth Growth
Stocks $ (500) $ 700 $ 2,200
Bonds $ (100) $ 600 $ 900
CDs $ 300 $ 500 $ 750
Mixture $ (200) $ 650 $ 1,300
Annual payoffs for an investment of $10,000
Criteria for Decision Making
Under Uncertainty
• Maximax payoff: Choose the best of the
best
• Maximin payoff: Choose the best of the
worst
• Hurwicz payoff: Use a weighted average of
the extremes
• Minimax regret: Minimize the maximum
opportunity loss
Maximax Criterion
1. Identify the maximum payoff for each alternative.
2. Choose the alternative with the largest maximum.
Slow Rapid
Stagnant Growth Growth Maximum
Stocks $ (500) $ 700 $ 2,200 $ 2,200
Bonds $ (100) $ 600 $ 900 $ 900
CDs $ 300 $ 500 $ 750 $ 750
Mixture $ (200) $ 650 $ 1,300 $ 1,300
Maximin Criterion
1. Identify the minimum payoff for each alternative.
2. Choose the alternative with the largest minimum.
Slow Rapid
Stagnant Growth Growth Minimum
Stocks $ (500) $ 700 $ 2,200 $ (500)
Bonds $ (100) $ 600 $ 900 $ (100)
CDs $ 300 $ 500 $ 750 $ 300
Mixture $ (200) $ 650 $ 1,300 $ (200)
Hurwicz Criterion
1. Identify the maximum payoff for each alternative.
2. Identify the minimum payoff for each alternative.
3. Calculate a weighted average of the maximum and the minimum using
and (1 - ) for weights.
4. Choose the alternative with the largest weighted average.
Slow Rapid =.7 =.3 Weighted
Stagnant Growth Growth Maximum Minimum Average
Stocks $ (500) $ 700 $ 2,200 $ 2,200 $ (500) $ 1,390
Bonds $ (100) $ 600 $ 900 $ 900 $ (100) $ 600
CDs $ 300 $ 500 $ 750 $ 750 $ 300 $ 615
Mixture $ (200) $ 650 $ 1,300 $ 1,300 $ (200) $ 850
Decision Alternatives
for Various Values of
Stocks Bonds CDs Mixture
Max Min Max Min Max Min Max Min
1- 2,200 -500 900 -100 750 300 1,300 -200
0.0 1.0 -500 -100 300 -200
0.1 0.9 -230 0 345 -50
0.2 0.8 40 100 390 100
0.3 0.7 310 200 435 250
0.4 0.6 580 300 480 400
0.5 0.5 850 400 525 550
0.6 0.4 1120 500 570 700
0.7 0.3 1390 600 615 850
0.8 0.2 1660 700 660 1000
0.9 0.1 1930 800 705 1150
1.0 0.0 2200 900 750 1300
Graph of Hurwicz Criterion
Selections for Various Values of
2500
Stocks
2000
1500 Mixture
1000 Bonds
CDs
500
0
-500 0.0 0.2 0.4 0.6 0.8 1.0
Investment Example:
Selected Regrets
I invested in stocks, and
Slow Rapid the economy grew slowly.
Stagnant Growth Growth I have no regrets.
Stocks $ (500) $ 700 $ 2,200
Bonds $ (100) $ 600 $ 900
CDs $ 300 $ 500 $ 750
Mixture $ (200) $ 650 $ 1,300
I invested in stocks.
Then the economy
I invested in CDs. stagnated. I regret not
Then the economy investing in CDs. I am
grew rapidly. I am $800 down from where
out $1,450. I could have been.
Investment Example:
Opportunity Loss Table
Slow Rapid
Stagnant Growth Growth
Stocks 800 0 0
Bonds 400 100 1,300
CDs 0 200 1,450
Mixture 500 50 900
Investment Example:
Calculating Opportunity Loss
Payoff Table Opportunity Loss Table
Slow Rapid Slow Rapid
Stagnant Growth Growth Stagnant Growth Growth
Stocks $ (500) $ 700 $ 2,200 Stocks 800 0 0
Bonds $ (100) $ 600 $ 900 Bonds 400 100 1,300
CDs $ 300 $ 500 $ 750 CDs 0 200 1,450
Mixture $ (200) $ 650 $ 1,300 Mixture 500 50 900
OLi,j = Max(column j) - Pi,j
Minimax Regret
1. Identify the maximum regret for each alternative.
2. Choose the alternative with the least maximum regret.
Slow Rapid
Stagnant Growth Growth Maximum
Stocks 800 0 0 800
Bonds 400 100 1,300 1,300
CDs 0 200 1,450 1,450
Mixture 500 50 900 900
Decision Making under Risk
• Probabilities of the states of nature have been
determined
– Decision-Making under uncertainty: probabilities of
the states of nature are unknown
– Decision-Making under risk: probabilities of the
states of nature are known (have been estimated)
• Decision Trees
• Expected Monetary Value of Alternatives
Decision Table with States of Nature
Probabilities for Investment Example
Slow Rapid
Stagnant Growth Growth
Probabilities
.25 .45 .30
Stocks $ (500) $ 700 $ 2,200
Bonds $ (100) $ 600 $ 900
CDs $ 300 $ 500 $ 750
Mixture $ (200) $ 650 $ 1,300
Decision Tree
for the Investment Example
Stagnant (.25) -$500
Chance Slow growth (.45)
$700
Node Rapid Growth (.30)
$2,200
Decision
Node Stocks Stagnant (.25)
-$100
Slow growth (.45)
Bonds Rapid Growth (.30) $600
$900
Stagnant (.25)
$300
CDs Slow growth (.45)
$500
Rapid Growth (.30)
Mixture $750
Stagnant (.25)
-$200
Slow growth (.45)
$650
Rapid Growth (.30)
$1,300
Expected Monetary Value Criterion
d X P
n
EMV i i, j j
j 1
where: d i
= decision alternative i
P j
= the probability of state j
Xi, j
= the payoff for decision i in state j
EMV Calculations
for the Investment Example
EMV stocks .25 500 .45 700 .3 2200 850
EMV bonds .25 100 .45 600 .30 900 515
EMV CDs .25 300 .45 500 .30 750 525
EMV mixture .25 200 .45 650 .30 1300 632.50
Decision Tree with Expected Monetary
Values for the Investment Example
Stagnant (.25) -$500
$850
Slow growth (.45)
$700
Rapid Growth (.30)
$2,200
Stagnant (.25)
Stocks -$100
$515 Slow growth (.45)
Bonds Rapid Growth (.30) $600
$900
Stagnant (.25)
$300
$525
CDs Slow growth (.45)
$500
Rapid Growth (.30)
Mixture $750
Stagnant (.25)
$623.50 -$200
Slow growth (.45)
$650
Rapid Growth (.30)
$1,300
EMV Criterion
for the Investment Example
1. Calculate the expected monetary value of each alternative.
2. Choose the alternative with the largest EMV.
Expected
Slow Rapid Monetary
Stagnant Growth Growth Value
0.25 0.45 0.30
Stocks $ (500) $ 700 $ 2,200 $ 850.00
Bonds $ (100) $ 600 $ 900 $ 515.00
CDs $ 300 $ 500 $ 750 $ 525.00
Mixture $ (200) $ 650 $ 1,300 $ 632.50
Expected Monetary Payoff with Perfect
Information for the Investment Example
Expected Monetary Payoff with Perfect Information
= ($300)(.25) + ($700)(.45) + ($2200)(.30)
= $1050
Slow Rapid
Stagnant Growth Growth
.25 .45 .30
Stocks $ (500) $ 700 $ 2,200
Bonds $ (100) $ 600 $ 900
CDs $ 300 $ 500 $ 750
Mixture $ (200) $ 650 $ 1,300
Expected Value of Perfect Information
for the Investment Example
Expected Value of Perfect Information
= Expected Monetary Payoff with Perfect Information - Max(EMV[d i])
= $1050 - $850
= $200
Utility
• The degree of pleasure or displeasure a
decision-maker has in being involved in the
outcome selection process given the risks
and opportunities available
– Risk-Avoider
– Risk-Neutral
– Risk-Taker
Risk Neutral: Indifferent to Owning “a”
or “b”
.5
$100,000
a
.5
-$0
b
$50,000
00
Risk Avoider: Indifferent to Owning “a”
or “b”
.5
$100,000
a
.5
-$0
b
$20,000
00
Risk Taker: Indifferent to Owning “a” or
“b”
.5
$100,000
a
.5
-$0
b
$70,000
00
Utility Curves for Game Players
Risk-Avoider
Chance of
Winning
the Contest
Risk
Neutral
Risk-Taker
Monetary Payoff
Revising Probabilities
in Light of Sample Information
• Bayes’ Rule
• Expected Value of Sample Information
Decision Table
for Investment Problem
No Rapid
Growth Growth
(.65) (.35)
Bonds $ 500 $ 100
Stocks $ (200) $ 1,100
Expected Monetary Value Criterion
for the Investment Example
Expected
No Rapid Monetary
Growth Growth Value
0.65 0.35
Bonds $ 500 $ 100 $ 360.00
Stocks $ (200) $ 1,100 $ 255.00
Decision Tree
for the Investment Example
No Growth (.65)
$500
EMV=$360
Bonds Rapid Growth (.35)
$360 $100
No Growth (.65)
-$200
Stocks
Rapid Growth (.35)
EMV=$255
$1,100
Historical Performance
of Economic Forecaster
Actual State of Economy
No Growth Rapid Growth
(s1) (s2)
Forecaster Predicts
No Growth (F1) .80 .30
Forecaster Predicts
Rapid Growth (F2) .20 .70
P(Fi|sj)
Bayes’ Rule
P(Y | Xi ) P( Xi )
P( Xi| Y )
P(Y | X 1) P( X 1) P(Y | X 2 ) P( X 2 ) P(Y | Xn ) P( Xn )
Revision Based on a Forecast
of No Growth (F1)
State of Prior Conditional Joint Revised
Economy Probabilities Probabilities Probabilities Probabilities
P(sj|F1)
No
Growth P(s 1) = .65 P(F 1| s 1) = .80 P(F 1 s 1) = .520 .520/.625 = .832
(s 1)
Rapid
Growth P(s 2) =.35 P(F 1| s 2) = .30 P(F 1 s 2) = .105 .105/.625 = .168
(s 2)
P(F1) = .625
Revision Based on a Forecast
of Rapid Growth (F2)
State of Prior Conditional Joint Revised
Economy Probabilities Probabilities Probabilities Probabilities
P(sj|F2)
No
Growth P(s 1) = .65 P(F 2| s 1) = .20 P(F 2 s 1) = .130 .130/.375 = .347
(s 1)
Rapid
Growth P(s 2) =.35 P(F 2| s 2) = .70 P(F 2 s 2) = .245 .245/.375 = .653
(s 2)
P(F1) = .375
Decision Tree for the Investment Example
After Revision of Probabilities
No Growth (.832)
$500
$432.80
Bonds Rapid Growth (.168)
$432.80 $100
Forecast No Growth (.832)
No Growth -$200
(.625)
Stocks
Rapid Growth (.168)
$18.40
$513.84 $1,100
Buy
Forecast No Growth (.347)
$500
$238.80
Bonds Rapid Growth (.653)
Forecast
Rapid Growth $648.90 $100
(.375)
No Growth (.347)
-$200
Stocks
Rapid Growth (.653)
$648.90
$1,100
Expected Value of Sample Information
for the Investment Example
Expected value of sample information
= expected monetary value with information
- expected monetary value without information
= $513.84 - $360
= $153.84
