Solutions to Practice Problems for Decision Analysis

1.
a) Max(A1) = 6, Max(A2) = 4, Max(A3) = 8. Maximax = 8 with alternative A3.
b) Min(A1) = 2, Min(A2) = 3, Min(A3) = 1. Maximin = 3 with alternative A2.
c) Construct the regret table (opportunity loss table)

STATE OF NATURE
ALTERNATIVE S1 S2 S3 Maximum
regret
A1 2 2 1 2
A2 5 0 2 5
A3 0 3 0 3
Alternative A1 has the minimum maximum regret.
2. a)
State of Nature
Alternative Demand is 10 Demand is 11 Demand is 12 Demand is 13

Buy 10 cases $50 $50 $50 $50

Buy 11 cases $47 $55 $55 $55
Buy 12 cases $44 $52 $60 $60
Buy 13 cases $41 $49 $57 $65
Prior Probability 0.2 0.4 0.3 0.1

b) Max(Buy 10) = $50, Max(Buy 11) = $55, Max(Buy 12) = $60, Max(Buy 13) = $65.
Maximax = $65 with buying 13 cases.
c) Min(Buy 10) = $50, Min(Buy 11) = $47, Min(Buy 12) = $44, Min(Buy 13) = $41.
Maximin = $50 with buying 10 cases.
d) Construct the regret table (opportunity loss table)

State of Nature
Alternative Sell 10 cases Sell 11 cases Sell 12 cases Sell 13 cases Maximum
regret
Buy 10 cases 0 5 10 15 15
Buy 11 cases 3 0 5 10 10
Buy 12 cases 6 3 0 5 6
Buy 13 cases 9 6 3 0 9

Buying 12 cases would minimize the maximum regret.

 e)
A B C D E F G H
1 Purchase Price $3
2 Selling Price $8
3
4 Payoff Table State of Nature
5 Sell Sell Sell Sell Expected
6 Alternative 10 11 12 13 Payoff
7 Buy 10 $50 $50 $50 $50 $50.00
8 Buy 11 $47 $55 $55 $55 $53.40
9 Buy 12 $44 $52 $60 $60 $53.60
10 Buy 13 $41 $49 $57 $65 $51.40
11
12 Prior Probability 0.2 0.4 0.3 0.1

Jean should buy 12 cases. The maximum expected payoff is $53.60.

3.
a) Min(South Korea) = 15.2, Min(China) = 17.6, Min(Taiwan) = 14.9, Min(Philippines) = 13.8,
Min(Mexico) = 12.5. Minimin = 12.5 with Mexico as the best alternative.
b) Max(South Korea) = 21.7, Max (China) = 19.0, Max (Taiwan) = 19.2, Max (Philippines) = 22.5,
Max(Mexico) = 25.0. Minimax = 19 with China as the best alternative.
c) Construct the regret table (opportunity loss table)

ECONOMIC AND
POLITICAL CLIMATES
COUNTRY Decline Same Improve Maximum
regret
SOUTH 2.7 2.3 2.7 2.7
KOREA
CHINA 0 1.7 5.1 5.1
TAIWAN 0.2 0.3 2.4 2.4
PHILIPPINES 3.5 0 1.3 3.5
MEXICO 6.0 4.4 0 6.0

Investing in Taiwan would minimize the maximum regret.

4. a)
State of Nature (Weather)
Alternative Dry Moderate Damp
Crop 1 $20 $35 $40
Crop 2 22.5 30 45
Crop 3 30 25 25
Crop 4 20 20 20
Prior Probability 0.3 0.5 0.2
 b) Crop 1 has the highest expected payoff of $31,500.

Note that Crop 4 is a dominated alternative, that has worse payoffs than Crop 3 in all situations. It
can be excluded from further analysis (but we left it in the tree anyway to practice the calculations).
5. a) Alternative A1 has the highest expected payoff of $35.
A B C D E F
1 Payoff Table State of Nature Expected
2 Alternative S1 S2 S3 Payoff
3 A1 $50 $100 -$100 $35
4 A2 $0 $10 -$10 $1
5 A3 $20 $40 -$40 $14
6
7 Prior Probability 0.5 0.3 0.2

b) With perfect information, choose A1 for when the state is S1, A1 when the state is S2, and A2
when the state is S3.

EP(with perfect information) = (0.5)($50) + (0.3)($100) + (0.2)(–$10) = $53

EVPI = EP(with perfect information) – EP (without more information)

= $53 – $35 = $18
c) Alternative A1 has the lowest Expected Opprotunity Loss (EOL) which is equal to EVPI=18.

Regret Table State of Nature Expected

Alternative S1 S2 S3 Opportunity Loss
A1 0 0 90 18
A2 50 90 0 52
A3 30 60 30 39
Prior Probability 0.5 0.3 0.2

d) Betsy should consider spending up to $18 to obtain more information.

 6.

Optimal policy: at Node 1, go up (choose the chance node with expectation 820)
If you make it to node 2 (only 20% of the time), then go down at node 2. (choose the sure
payoff of 900)
 7.

Optimal Policy: at the first decision node, go down (choose the chance node with expectation 8).

If you get to the 2nd decision node in the bottom sub-tree (40% of the time), go down
(choosing the chance node with expectation 5).
8. (Note: this decision tree continues on the next page.)
The comptroller should invest in stocks the first year. If there is growth during the first year then
she should invest in stocks again the second year. If there is a recession during the first year then
she should invest in bonds for the second year. The expected payoff is $122.94 million.
9. a through d)
(0.1)(0.95)=0.095 .095/0.14 = 0.679
0.95 User & Pos User | Pos
Pos | User

0.05
(0.1)(0.05)=0.005 .005/0.86 = 0.006
0.1 Neg | User
User User & Neg User | Neg

0.9 (0.9)(0.05)=0.045 .045/0.14 = 0.321

0.05 Nonuser & Pos Nonuser | Pos
Nonuser
Pos | Nonuser

0.95
Neg | Nonuser (0.9)(0.95)=0.855 .855/0.86 = 0.994
Nonuser & Neg Nonuser | Neg
a) P(User|Pos)=0.679
b) P(Nonuser|Pos)=0.321
c) P(User|Neg)=0.006
d) P(Nonuser|Neg)=0.994

Altenatively you can directly use Bayes’ theorem to answer parts a to d. First we need to write all
probabilities in mathmatical format:
P(Pos|User)=0.95, P(Neg|User)=1- P(Pos|User)=0.05
P(Neg|NonUser)=0.95, P(Pos|NonUser)=1- P(Neg|NonUser)=0.05
P(User)=0.1, P(NonUser)=1-0.1=0.9

Now we can calculate all posterior probabilities. For example, for part a, the probability that the
athlete is a drug user, given that the test is positive: P(User|Pos)

P(Pos|User) × P(User)
P(User|Pos) =
P(Pos|User) × P(User) + P(Pos|NonUser) × P(NonUser)

0.95 × 0.1
P(User|Pos) = = 0.679
0.95 × 0.1 + 0.05 × 0.9
10. a)
State of Nature
Alternative Successful Unsuccessful
Develop new product $1,500,000 –$1,800,000
Don’t develop new product 0 0
Prior Probabilities 0.667 0.333

b) Choosing to develop the product maximizes the expected payoff ($400,000).

A B C D E
1 Payoff Table ($millions) Expected
2 State of Nature Payoff
3 Alternative Successful Unsuccessful ($millions)
4 Develop Product 1.5 -1.8 0.4
5 Don't Develop Product 0 0 0
6
7 Prior Probability 0.667 0.333

c) With perfect information, Telemore should develop the product if it would be successful,
and don’t if it will be unsuccessful.

EP(perfect information) = (0.667)(1.5) + (0.333)(0) = $1 million.

EVPI = EP(with perfect information) – EP(without more information)

= $1,000,000 – $400,000 = $600,000.

We know that EVSI < EVPI, so a could market survey could be worth up to
$600,000 in additional value. Since the market survey only costs $100,000, this
indicates that we should consider doing the market survey. So, we should now do
the Bayesian updating and build the tree:
d)
B C D E F G H
3 Data: P(Finding | State)
4 State of Prior Finding
5 Nature Probability Predict Successful Predict Unsuccessful
6 Successful 0.667 0.8 0.2
7 Unsuccessful 0.333 0.3 0.7
8
9
10
11
12 Posterior P(State | Finding)
13 Probabilities: State of Nature
14 Finding P(Finding) Successful Unsuccessful
15 Predict Successful 0.633 0.842 0.158
16 Predict Unsuccessful 0.367 0.364 0.636
17
18
19
 e) They should conduct the survey, and develop the product if the survey predicts the product will
be successful. If the survey predicts the product will be unsuccessful, they should not develop the
product. The expected payoff after the test cost is $620,000-100,000=520,000.

f)

EVSI = EP(with survey information) – EP(without more information)

= $620,000 – $400,000 = $220,000.
11. a)
 B C D E F G H
3 Data: P(Finding | State)
4 State of Prior Finding
5 Nature Probability Pass Test Fail Test
6 Successful 0.7 0.9 0.1
7 Not Successful 0.3 0.1 0.9
8
9
10
11
12 Posterior P(State | Finding)
13 Probabilities: State of Nature
14 Finding P(Finding) Successful Not Successful
15 Pass Test 0.6600 0.9545 0.0455
16 Fail Test 0.3400 0.2059 0.7941
17
18
19

The optimal policy is not to pay for testing and to hire Matthew.
b) If the fee is less than 1,242,000-1,220,000=$22,000 then the testing is worthwhile.