Problems 1. Here is a payoff matrix; parentheses designate a loss. The probabilities of the states of nature, p(1) and p(2), are unknown. | | States of Nature | | Alternatives | 1 | 2 | | A B C | $500 700 (300) | $300 100 600 |
- Which alternative should be selected under the maximax criterion?
- Which alternative should be selected under the maximin criterion?
- Which alternative should be selected under the Laplace criterion?
- What is the opportunity loss table?
- Which alternative should be selected under the minimax regret criterion?
2. Refer to Problem 1. Construct the sensitivity graph, as shown on
p. 207 of your textbook. - For what range of values of p(2) is alternative A preferred?
- For what range of values of p(2) is alternative B preferred?
- For what range of values of p(2) is alternative C preferred?
3. Fall has come, and the department manager of the Purple Pentagon Discount Store must decide how many snow blowers to order for the winter season. He believes there is an association between the weather and the demand for snow blowers: a winter with heavy snow should produce a high demand, and a winter with light snow should produce a low demand. However, if he guesses wrong and orders a large number of snow blowers when the weather turns out to have light snow, he will suffer a loss. Here is his payoff matrix. | States of Nature | | Alternatives | Heavy snow | Medium snow | Light snow | | Large Order Medium Order Small Order | $5,000 2,000 500 | $1,000 3,000 500 | $(4,000) (2,000) 1,000 |
- Which alternative should be selected under the maximax criterion?
- Which alternative should be selected under the maximin criterion?
- Which alternative should be selected under the Laplace criterion?
- What is the opportunity loss table?
- Which alternative should be selected under the minimax regret criterion?
4. Refer to Problem 3. Suppose that the following probabilities are available
regarding the snowfall next winter:
p(heavy snow) = .10
p(medium snow) = .50
p(light snow) = .40 - Which alternative should be selected under the expected value criterion?
- Which alternative should be selected under the lowest expected loss criterion?
- What is the value of the EVPI?
- What is the interpretation of the EVPI?
5. Solve problem #4 by using an EXCEL spreadsheet. - Find the expected monetary values.
- Find the largest expected monetary value.
- Find the expected value of perfect information.
- Compare the answers on your spreadsheet to the answers to problem #4.
6. (One step beyond) The telephone company has three billing plans. - An unlimited number of local telephone calls per month for $20.00
- Fifty (or fewer) local telephone calls per month for $16.00 plus $0.08
per call above 50 calls.
- Twenty (or fewer) local telephone calls per month for $12.00 plus $0.10
per call above 20 calls.
The subscriber is uncertain as to whether the number of local telephone calls
will be about 60 per month, or about 90 per month, or about 120 per month, or
about 160 per month. Use decision theory to determine which billing plan the subscriber
should select:
- What is the payoff table?
- Which plan should be selected under the minimax regret criterion?
- Why is it improper to use the maximax criteria here?
- Suppose that the subscriber assesses probabilities as follows:
p(60 calls) = .40 p(120 calls) = .20
p(90 calls) = .30 p(160 calls) = .10
Which plan should be selected under the best expected value criterion? - Which plan should be selected under the least expected regret criterion?
Solutions 1. a. The maximax criterion involves selecting the best payoff for each alternative and then selecting the best payoff in that set. | Alternative | Best Payoff | | A B C | $500 700 (best) 600 |
Implement Alternative B. b. The maximin criterion involves selecting the worst payoff for each alternative and then selecting the best payoff in that set. | Alternative | Worst Payoff | | A B C | $300 (best) 100 (300) |
Implement Alternative A. c. The LaPlace criterion involves selecting the best mean payoff. | Alternative | Mean Payoff | | A B C | $400* (tie) 400 (tie) 150 |
Implement Alternative A or Alternative B. *Note: (500 + 300)/2 = 400. d. Find the regret for each cell by subtracting each entry in the
state of nature column from the
maximum entry in the column. | State of Nature | | Alternative | 1 | 2 | | A B C | 200 0 1,000 | 300 500 0 |
e. The minimax regret criterion involves selecting the worst regret for each alternative, and then selecting the best of that set. | Alternative | Worst Regret | | A B C | $300 (best) 500 1,000 |
Implement Alternative A.
At the points where the lines in the diagram intersect the y coordinates
are equal; y = a + bx, where a is the intercept on the (1) column and b
= payoff #2 - payoff #1. A line: y = 500 + (300 - 500)x = 500 - 200x.
B line: y = 700 + (100 - 700)x = 700 - 600x.
C line: y = -300 + (600 - (-300))x = -300 + 900x. The A line and the B line intersect. 500 - 200p(2) = 700 - 600p(2)
-200 = -400p(2)
p(2) = 0.5
p(1) = 1 - p(2) = 1 - 0.5 = 0.5. The A line and the C line intersect. 500 - 200p(2) = -300 + 900p(2)
800 = 1100p(2)
p(2) = 0.727
p(1) = 1 - p(2) = 1 - 0.727 = .273.
a. Choose A for 0.5 < p(2) < 0.727.
b. Choose B for 0 < p(2) < 0.5.
c. Choose C for 0.727 < p(2) < 1.0. 3. a. | Alternative | Best Payoff | | Large Order Medium Order Small Order | $5,000 (best) 3,000 1,000 |
Place a large order. b. | Alternative | Worst Payoff | | Large Order Medium Order Small Order | ($5,000) (2,000) 500 (best) |
Place a small order. c. | Alternative | Mean Payoff | | Large Order Medium Order Small Order | $667* 1,000 667 |
Place a medium order. *Note: (5000 + 1000 - 4000)/3 = 667. d. Find the regret for each cell in the payoff matrix by subtracting each entry in the state of nature column from the maximum payoff in the column. For the Heavy Snow column, the best payoff is $5,000, and the regrets are 0, $3,000, and $4,500. | | State of Nature | | Alternative | Heavy Snow | Medium Snow | Light Snow | | Large Order Medium Order Small Order | 0 3,000 4,500 | 2,000 0 2,500 | 5,000 3,000 0 |
e. | Alternative | Worst Regret | | Large Order Medium Order Small Order | $5,000 3,000 (best) 4,500 |
Place a medium order. *Note: (5000 + 1000 - 4000)/3 = 667. d. Find the regret for each cell in the payoff matrix by subtracting each entry in the state of nature column from the maximum payoff in the column. For the Heavy Snow column, the best payoff is $5,000, and the regrets are 0, $3,000, and $4,500. | | State of Nature | | Alternative | Heavy Snow | Medium Snow | Light Snow | | Large Order Medium Order Small Order | 0 3,000 4,500 | 2,000 0 2,500 | 5,000 3,000 0 |
e. | Alternative | Worst Regret | | Large Order Medium Order Small Order | $5,000 3,000 (best) 4,500 |
Place a medium order. *Note: 5000(.10) + 1000(.50) - 4000(.40) = -600. b. | | State of Nature | | | Alternative | Heavy Snow .10 | Medium Snow .50 | Light Snow .40 | Expected Loss | | Large Order Medium Order Small Order | 0 3,000 4,500 | 2,000 0 2,500 | 5,000 3,000 0 | 3,000* 1,500 (best) 1,700 |
*Note: 0(.10) + 2000(.50) + 5000(.40) = 3000.
Place a medium order. The strategy which is selected by the least expected
loss criterion
will always be the same as the one selected by the best expected value criterion. c. EVPI = least expected loss = $1,500.
d. $1,500 is the absolute maximum which might be spent to obtain information
about the states
of nature. The manager might, for example, hire a weather forecaster to
forecast next
winter's snowfall. | | State of Nature | | Alternative | Heavy Snow .10 | Medium Snow .50 | Light Snow .40 | | Large Order Medium Order Small Order | $5,000 2,000 500 | $1,000 3,000 500 | -$4,000 2,000 1,000 |
5. Image179 (21.0K)Image179 - In Cell H-8: =D6*D8+E6*E8+F6*F8
In Cell H-9: =D6*D9+E6*E9+F6*F9
In Cell H-10: =D6*D10+E6*E10+F6*F10- In Cell D-12: =MAX(H8:H10)
- In Cell D-13: =MAX(D8:D10)*D6+MAX(E8:E10)*E6+MAX(F8:F10)*F6-D12
- Did you get the same answers as in #4-a and #4-c?
6. a. It is necessary to estimate the total cost of each possible
number of calls under each billing plan. Plan a) The cost will be $20, regardless of the number of calls. Plan b) The cost will be $16 + (number of calls - 50) x $0.08, when
there are more than 50 calls.
Plan c) The cost will be $12 + (number of calls - 20) x $0.10, when
there are more than 20 calls. The payoff table will be: | | State of Nature | | Alternative | 60 calls | 90 calls | 120 calls | 160 calls | | Plan a Plan b Plan c | 20.00 16.80 16.00 | 20.00 19.20 19.00 | 20.00 21.60 22.00 | 20.00 24.80 26.00 |
b. Select the best payoff for each state of nature; they are 16.00,
19.00,20.00, and 20.00. Subtract every entry in the column from the best entry. The
differences are the regrets. | | State of Nature | | Alternative | 60 calls | 90 calls | 120 calls | 160 calls | | Plan a Plan b Plan c | 4.00 .80 0 | 1.00 .20 0 | 0 1.60 2.00 | 0 4.80 6.00 |
Pick the worst regret for each alternative: Alternative Worst Regret Plan a 4.00 (best)
Plan b 4.80
Plan c 6.00
The subscriber should select Plan a. c. It is improper to use the maximax criterion here because the payoffs
are expenses rather than incomes.
d. Incorporate the probabilities into the payoff table and calculate
the expected value of each plan. | | State of Nature | | Probability | .40 | .30 | .20 | .10 | | Alternative | 60 calls | 90 calls | 120 calls | 160 calls | | Plan a *Plan b Plan c | 20.00 16.80 16.00 | 20.00 19.20 19.00 | 20.00 21.60 22.00 | 20.00 24.80 26.00 |
The subscriber should select Plan c.
*Note: (16.80)(.40) + (19.20)(.30) + (21.60)(.20) + (24.80)(.10) = 19.28. e. The subscriber should select Plan c. |