Midterm solutions

Question 1 :

a- F One can be desperate to get only the best over solution. Further it is not wrong to be
extremely optimistic
b- T PI is always an upper bound of imperfect information as the best/worst decisions
outcomes are multiplied by 1/0 instead of lower probability values of the posterior
distribution.
c- T Yes. In fact, the EMV is a good average when the fluctuations are low.
d- F As in one, no rule to say that one criterion is better than another.
e- F. In fact, EVII may be zero yielding no value at all.

Question 2 :

1- a-

E1 E2 E3 E4 Max-max
A1 700000 150000 10000 -55000 700000
A2 1000000 200000 -100000 -140000 1000000
A3 800000 100000 50000 0 800000

a- Based on the max-max criterion, A2 is the best alternative that yields 1000000DT.
b- Savage Min-Max criterion

E1 E2 E3 E4 Max
A1 300000 50000 40000 55000 300000
A2 0 0 150000 140000 150000
A3 200000 100000 0 0 200000

Based on Min-Max regret criterion, the best alternative is A2.

c- Table

A1 700000α-55000(1- α) = 755000 α-55000

1
 A2 1000000α-140000(1- α) = 1140000 α-140000
A3 800000α

A1 is always dominated by A3

A3 is the best alternative if 800000α ≥ 1140000 α-140000. Therefore, 140000≥340000 α. Hence,

α≤0.42

Thus, A3 is the best alternative if α≤0.42. Otherwise, A2 is the best solution.

2- a- A2 is the best alternative with an average of 312000TND in cash flows.

E1 (0.3) E2 (0.3) E3 (0.2) E4 (0.2) EMV

A1 700000 150000 10000 -55000 246000
A2 1000000 200000 -100000 -140000 312000
A3 800000 100000 50000 0 280000
b- Given that it is a one shot decision with highly fluctuating values, EMV is not recommended as
the alternatives are highly risky.

C-

Alternative Expected regret

A1 (300000*0.3)+(50000*0.3)+(40000*0.2)+(55000*0.2)=124000
A2 (150000*0.2)+(140000*0.2)=58000
A3 (200000*0.3)+(100000*0.3)=90000

A2 is the best alternative with an expected regret of 58000TND.

Question3:

1- The firm has three alternatives:

A1: go on time
A2: Request delay
A3: Abandon the project
2- The chance events of A1:

2
 E1: Obtain confirmed flights
E2: Do not obtain confirmed flights
E3: Win the project
E4: Loose the project
The chance events of A2:
E5: Delay request accepted
E6: Delay request not accepted
E7: Win the project
E8: Loose the project
3- Decision tree
Not confirmed flights (0.4)
-25 000
Go on time

90 200 Win the project (0.8)

215 000
90 200 Confirmed flights (0.6)

167 000 Loose the project (0.2)

-25 000

Win the project (0.6)

175 000
Request accepted (0.5)

Request delay 79 000 Loose the project (0.4)

-65 000

39 500
Request not accepted (0.5)
0

Abandon the project

0
4- Based on EMV, the optimal policy of the firm is to opt for going on time. This alternative
yields an expected profit of 90 200 TND.
5- The use of the EMV criterion is not suitable for this problem as it deals with a one shot
decion with high fluctuations.

3
 6- Bayes table
C (0.6) NC (0.4) 
“confirmed” 0.9 0.1 0.54 0.04 0.58 0.93 0.07
“not confirmed 0.1 0.9 0.06 0.36 0.42 0.14 0.86

Do not opt for the travel agency

90 200

W (0.8)
213 000
C/ « C » (0.93)
103 654,800

Go on time 165 000 L(0.2)

-27 000

151 560
NC/ « C » (0.07)
-27 000
« Confirmed » (0.58)

151 560 Request delay

37 500
Opt for the travel agency
Abandon the project
103 654,800 -2 000

W (0.8)
213 000

C/ « NC » (0.14)

Go on time 165 000 L(0.2)

-27 000
-120 000
NC/ « NC » (0.86)
-27 000
« Not confirmed » (0.42)

Request delay
37 500 37 500

Abandon the project

-2 000
4
Yes, the consulting firm should opt for the manager’s information as the expected profit will
jump to 103654,800 TND instead of 90 200 TND.
7- Decision tree

5
 No PI
90 200

Go on time
90 200
90 200 « Request accepted » (0.5)
Request delay
79 000
PI
90 200 Abandon the project
0
90 200

Go on time
90 200
« Request not accepted » (0.5)
Request delay
0

90 200
Abandon the project
0

EVPI = Expected profit with additional information – expected profit with no perfect information
EVPI= 90 200 – 90 200 = 0 TND.
The perfect information does not help to improve the decision policy.