Вы находитесь на странице: 1из 30

Chapter 19: Decision Analysis 1

Chapter 19 Decision Analysis

LEARNING OBJECTIVES Chapter 19 describes how to use decision analysis to improve management decisions, thereby enabling you to: 1. 2. 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.

3. 4. 5.

CHAPTER TEACHING STRATEGY The notion of contemporary decision making is built into the title of the text as a statement of the importance of recognizing that statistical analysis is primarily done as a decision-making tool. For the vast majority of students, statistics take on importance only in as much as they aid decision-makers in weighing various alternative pathways and helping the manager make the best possible determination. It has been an underlying theme from chapter 1 that the techniques presented should be considered in a decisionmaking context. This chapter focuses on analyzing the decision-making situation and presents several alternative techniques for analyzing decisions under varying conditions. Early in the chapter, the concepts of decision alternatives, the states of nature, and the payoffs are presented. It is important that decision makers spend time brainstorming about possible decision alternatives that might be available to them. Sometimes the best alternatives are not obvious and are not immediately considered. The international focus

Chapter 19: Decision Analysis 2 on foreign companies investing in the U.S. presents a scenario in which there are several possible alternatives available. By using cases such as the Fletcher-Terry caseTerry case at the chapter's end, students can practice enumerating possible decision alternatives. States of nature are possible environments within which the outcomes will occur over which we have no control. These include such things as the economy, the weather, health of the CEO, wildcat strikes, competition, change in consumer demand, etc. While the text presents problems with only a few states of nature in order to keep the length of solution reasonable, students should learn to consider as many states of nature as possible in decision making. Determining payoffs is relatively difficult but essential in the analysis of decision alternatives. Decision-making under uncertainty is the situation in which the outcomes are not known and there are no probabilities given as to the likelihood of them occurring. With these techniques, the emphasis is whether or not the approach is optimistic, pessimistic, or weighted somewhere in between. In making decisions under risk, the probabilities of each state of nature occurring are known or are estimated. Decision trees are introduced as an alternative mechanism for displaying the problem. The idea of an expected monetary value is that if this decision process were to continue with the same parameters for a long time, what would the long-run average outcome be? Some decisions lend themselves to long-run average analysis such as gambling outcomes or insurance actuary analysis. Other decisions such as building a dome stadium downtown or drilling one oil well tend to be more one time activities and may not lend themselves as nicely to expected value analysis. It is important that the student understand that expected value outcomes are long-run averages and probably will not occur in single instance decisions. Utility is introduced more as a concept than an analytic technique. The idea here is to aid the decision-maker in determining if he/she tends to be more of a risktaker, an EMV'r, or risk-averse. The answer might be that it depends on the matter over which the decision is being made. One might be a risk-taker on attempting to employ a more diverse work force and at the same time be more risk-averse in investing the company's retirement fund.

Chapter 19: Decision Analysis 3

CHAPTER OUTLINE 19.1 The Decision Table and Decision Making Under Certainty Decision Table Decision-Making Under Certainty 19.2 Decision Making Under Uncertainty Maximax Criterion Maximin Criterion Hurwicz Criterion Minimax Regret 19.3 Decision Making Under Risk Decision Trees Expected Monetary Value (EMV) Expected Value of Perfect Information Utility 19.4 Revising Probabilities in Light of Sample Information Expected Value of Sample Information

KEY TERMS Decision Alternatives Decision Analysis Decision Making Under Certainty Decision Making Under Risk Decision Making Under Uncertainty Decision Table Decision Trees EMV'er Expected Monetary Value (EMV) Expected Value of Perfect Information Expected Value of Sample Information Hurwicz Criterion Maximax Criterion Maximin Criterion Minimax Regret Opportunity Loss Table Payoffs Payoff Table Risk-Avoider Risk-Taker States of Nature Utility

Chapter 19: Decision Analysis 4

SOLUTIONS TO PROBLEMS IN CHAPTER 19 19.1 S1 d1 250 d2 110 d3 390 S2 175 100 140 S3 -25 70 -80 Max 250 110 390 Min -25 70 -80 decision: Select d3 decision: Select d2

a.) Max {250, 110, 390} = 390 b.) Max {-25, 70, -80} = 70 c.) For = .3 d1: .3(250) + .7(-25) = 57.5 d2: .3(110) + .7(70) = 82 d3: .3(390) + .7(-80) = 61 decision: Select d2 For = .8 d1: .8(250) + .2(-25) = 195 d2: .8(110) + .2(70) = 102 d3: .8(390) + .2(-80) = 296 decision: Select d3

Comparing the results for the two different values of alpha, with a more pessimist point-of-view ( = .3), the decision is to select d2 and the payoff is 82. Selecting by using a more optimistic point-of-view ( = .8) results in choosing d3 with a higher payoff of 296.

Chapter 19: Decision Analysis 5 d.) The opportunity loss table is: S1 d1 140 d2 280 d3 0 S2 0 75 35 S3 95 0 150 Max 140 280 150

The minimax regret = min {140, 280, 150} = 140 Decision: Select d1 to minimize the regret. 19.2 d1 d2 d3 S1 50 80 20 S2 70 20 45 85 -10 S3 120 75 30 -30 65 S4 110 100 60 -20 80 Max 120 100 60 100 80 Min 50 20 20 -30 -10

d4 100 d5 0

a.) Maximax = Max {120, 100, 60, 100, 80} = 120 Decision: Select d1 b.) Maximin = Max {50, 20, 20, -30, -10} = 50 Decision: Select d1 c.) = .5 Max {[.5(120)+.5(50)], [.5(100)+.5(20)], [.5(60)+.5(20)], [.5(100)+.5(-30)], [.5(80)+.5(-10)]}= Max { 85, 60, 40, 35, 35 } = 85 Decision: Select d1

Chapter 19: Decision Analysis 6

d.) Opportunity Loss Table:November 8, 1996 S1 d1 d2 d3 d4 50 20 80 0 S2 15 65 40 0 95 S3 0 45 90 150 55 S4 0 10 50 130 30 Max 50 65 90 150 100

d5 100

Min {50, 65, 90, 150, 100} = 50 Decision: Select d1 19.3 A B R 60 10 D 15 25 40 25 I -25 30 15 5 Max Min 60 30 40 25 -25 10 -10 5

C -10 D 20

Maximax = Max {60, 30, 40, 25} = 60 Decision: Select A Maximin = Max {-25, 10, -10, 5} = 10 Decision: Select B

Chapter 19: Decision Analysis 7

19.4 None Few Many

Not -50 -200 -600

Somewhat -50 300 100

Very -50 400 1000

Max -50 400 1000

Min -50 -200 -600

a.) For Hurwicz criterion using = .6: Max {[.6(-50) + .4(-50)], [.6(400) + .4(-200)], [.6(1000) + .4(-600)]} = {-50, -160, 360}= 360 Decision: Select "Many" b.) Opportunity Loss Table: Not None Few Many 0 150 550 Somewhat 350 0 200 Very 1050 600 0 Max 1050 600 550

Minimax regret = Min {1050, 600, 550} = 550 Decision: Select "Many"

Chapter 19: Decision Analysis 8

19.5, 19.6

Chapter 19: Decision Analysis 9 19.7 Expected Payoff with Perfect Information = 5(.15) + 50(.25) + 20(.30) + 8(.10) + 6(.20) = 31.75 Expected Value of Perfect Information = 31.25 - 25.25 = 6.50 19.8 a.) & b.)

c.) Expected Payoff with Perfect Information = 150(40) + 450(.35) + 700(.25) = 392.5 Expected Value of Perfect Information = 392.5 - 370 = 22.50

Chapter 19: Decision Analysis 10

19.9 Lock-In No

Down(.30) -150 175

Up(.65) 200 -250

No Change(.05) 0 0

EMV 85 -110

Decision: Based on the highest EMV)(85), "Lock-In" Expected Payoff with Perfect Information = 175(.30) + 200(.65) + 0(.05) = 182.5 Expected Value of Perfect Information = 182.5 - 85 = 97.5

19.10 No Layoff Layoff 1000 Layoff 5000

EMV -960 -320 400

Decision: Based on maximum EMV (400), Layoff 5000 Expected Payoff with Perfect Information = 100(.10) + 300(.40) + 600(.50) = 430 Expected Value of Perfect Information = 430 - 400 = 30

19.11 a.) EMV = 200,000(.5) + (-50,000)(.5) = 75,000 b.) Risk Avoider because the EMV is more than the investment (75,000 > 50,000) c.) You would have to offer more than 75,000 which is the expected value.

Chapter 19: Decision Analysis 11

19.12 a.) d1 d2

S1(.30) 350 -200

S2(.70) -100 325

EMV 35 167.5

Decision: Based on EMV, maximum {35, 167.5} = 167.5 b. & c.) For Forecast S1: Prior S1 S2 .30 .70 Cond. .90 Joint .27 Revised .6067 .3933

.25 .175 F(S1) = .445

For Forecast S2: Prior S1 S2 .30 .70 Cond. .10 Joint .030 Revised .054 .946

.75 .525 F(S2) = .555

Chapter 19: Decision Analysis 12

EMV with Sample Information = 241.63

d.) Value of Sample Information = 241.63 - 167.5 = 74.13

Chapter 19: Decision Analysis 13

19.13 Dec(.60) S M L -225 125 350 Inc(.40) 425 -150 -400 EMV 35 15 50

Decision: Based on EMV = Maximum {35, 15, 50} = 50 For Forecast (Decrease): Prior Decrease Increase .60 .40 Cond. .75 Joint .45 Revised .8824 .1176

.15 .06 F(Dec) = .51

For Forecast (Increase): Prior Decrease Increase .60 .40 Cond. .25 Joint .15 Revised .3061 .6939

.85 .34 F(Inc) = .49

Chapter 19: Decision Analysis 14

The expected value with sampling is 244.275 EVSI = EVWS - EMV = 244.275 - 50 = 194.275

Chapter 19: Decision Analysis 15

19.14 Don't Plant Small Large

Decline(.20) Same(.30) Increase(.50) 20 -90 -600 0 10 -150 -40 175 800

EMV -16 72.5 235

Decision: Based on Maximum EMV = Max {-16, 72.5, 235} = 235, plant a large tree farm For forecast decrease: Prior .20 .30 .50 Cond. .70 .02 .02 Joint .140 .006 .010 Revised .8974 .0385 .0641

P(Fdec) = .156 For forecast same: Prior .20 .30 .50 Cond. .25 .95 .08 Joint .05 .285 .040 Revised .1333 .7600 .1067

P(Fsame) = .375 For forecast increase: Prior .20 Cond. .05 Joint .01 Revised .0213

Chapter 19: Decision Analysis 16

.30 .50

.03 .90

.009 .45

.0192 .9595

P(Finc) = .469

Chapter 19: Decision Analysis 17

The Expected Value with Sampling Information is 360.413 EVSI = EVWSI - EMV = 360.413 - 235 = 125.413

19.15 Drill Don't Drill

Oil(.11) 1,000,000 0

No Oil(.89) -100,000 0

EMV 21,000 0

Decision: The EMV for this problem is Max {21,000, 0} = 21,000. The decision is to Drill. Actual Oil No Oil Oil Forecast No Oil Forecast Oil: State Oil No Oil Prior .11 .89 Cond. .20 .10 Joint .022 .089 Revised .1982 .8018 .80 .90 .20 .10

P(FOil) = .111 Forecast No Oil: State Oil No Oil Prior .11 .89 Cond. .80 .90 Joint .088 .801 Revised .0990 .9010

P(FNo Oil) = .889

Chapter 19: Decision Analysis 18

The Expected Value With Sampling Information is 21,012.32 EVSI = EVWSI - EMV = 21,000 - 21,012.32 = 12.32

Chapter 19: Decision Analysis 19

19.16 d1 d2 d3 d4

S1 50 -75 25 75

S2 100 200 40 10

Max. 100 200 40 75

Min. 50 -75 25 10

a.) Maximax:

Max {100, 200, 40, 75} = 200

Decision: Select d2 b.) Maximin: Max {50, -75, 25, 10} = 50

Decision: Select d1 c.) Hurwicz with = .6 d1: d2: d3: d4: 100(.6) + 50(.4) = 80 200(.6) + (-75)(.4) = 90 40(.6) + 25(.4) = 34 75(.6) + 10(.4) = 49

Max {80, 90, 34, 49} = 90 Decision: Select d2 d.) Opportunity Loss Table: S1 d1 d2 d3 d4 25 150 50 0 S2 100 0 160 190 Maximum 100 150 160 190

Min {100, 150, 160, 190} = 100 Decision: Select d1

Chapter 19: Decision Analysis 20

19.17

b.) d1: 400(.3) + 250(.25) + 300(.2) + 100(.25) = 267.5 d2: 300(.3) + (-100)(.25) + 600(.2) + 200(.25) = 235 Decision: Select d1 c.) Expected Payoff of Perfect Information: 400(.3) + 250(.25) + 600(.2) + 200(.25) = 352.5 Value of Perfect Information = 352.5 - 267.5 = 85

Chapter 19: Decision Analysis 21

19.18 d1 d2 d3

S1(.40) 200 -75 175

S2(.60) 150 450 125

EMV 170 240 145

Decision: Based on Maximum EMV = Max {170, 240, 145} = 240 Select d2 Forecast S1: State S1 S2 Prior .4 .6 Cond. .9 .3 Joint .36 .18 Revised .667 .333

P(FS1) = .54 Forecast S2: State S1 S2 Prior .4 .6 Cond. .1 .7 Joint .04 .42 Revised .087 .913

P(FS2) = .46

Chapter 19: Decision Analysis 22

The Expected Value With Sample Information is 285.00 EVSI = EVWSI - EMV = 285 - 240 = 45

Chapter 19: Decision Analysis 23

19.19 Small Modest Large

Small 200 100 -300

Moderate 250 300 400

Large 300 600 2000

Min 200 100 -300

Max 300 600 2000

a.) Maximax: Max {300, 600, 2000} = 2000 Decision: Large Number Minimax: Max {200, 100, -300} = 200 Decision: Small Number b.) Opportunity Loss: Small Small Modest Large 0 100 500 Moderate 150 100 0 Large 1700 1400 0 Max 1700 1400 500

Min {1700, 1400, 500} = 500 Decision: Large Number c.) Minimax regret criteria leads to the same decision as Maximax.

Chapter 19: Decision Analysis 24

19.20 Low Medium High a.) = .1:

No -700 -300 100

Low -400 -100 125

Fast 1200 550 150

Max 1200 550 150

Min -700 -300 100

Low: 1200(.1) + (-700)(.9) = -510 Medium: 550(.1) + (-300)(.9) = -215 High: 150(.1) + 100(.9) = 105 Decision: Price High (105) b.) = .5: Low: 1200(.5) + (-700)(.5) = 250 Medium: 550(.5) + (-300)(.5) = 125 High: 150(.5) + 100(.5) = 125 Decision: Price Low (250) c.) = .8: Low: 1200(.8) + (-700)(.2) = 820 Medium: 550(.8) + (-300)(.2) = 380 High: 150(.8) + 100(.2) = 140 Decision: Price Low (820) d.) Two of the three alpha values (.5 and .8) lead to a decision of pricing low. Alpha of .1 suggests pricing high as a strategy. For optimists (high alphas), pricing low is a better strategy; but for more pessimistic people, pricing high may be the best strategy.

Chapter 19: Decision Analysis 25

19.21 Reg. Weekend Not Open

Mild(.75) 2000 1200 -300

Severe(.25) -2500 -200 100

EMV 875 850 -200

Decision: Based on Max EMV = Max{875, 850, -200} = 875, open regular hours.

Expected Value with Perfect Information = 2000(.75) + 100(.25) = 1525 Value of Perfect Information = 1525 - 875 = 650

Chapter 19: Decision Analysis 26

19.22 Don't Produce Produce

Weaker(.35) Same(.25) Stronger(.40) -700 1800 -200 400 150 -1600

EMV -235 90

Decision: Based on Max EMV = Max {-235, 90} = 90, select Produce. Expected Payoff With Perfect Information = 1800(.35) + 400(.25) + 150(.40) = 790 Value of Perfect Information = 790 - 90 = 700

Chapter 19: Decision Analysis 27 19.23 Automate Do Not Red.(.15) -40,000 5,000 Con.(.35) -15,000 10,000 Inc.(.50) 60,000 -30,000 EMV 18,750 -10,750

Decision: Based on Max EMV = Max {18750, -10750} = 18,750, Select Automate Forecast Reduction: State R C I Prior .15 .35 .50 Cond. Joint Revised .60 .2333 .1667

.60 .09 .10 .035 .05 .025 P(FRed) = .150

Forecast Constant: State R C I Prior .15 .35 .50 Cond. Joint Revised .10 .6222 .2778

.30 .045 .80 .280 .25 .125 P(FCons) = .450

Forecast Increase: State R C I Prior .15 .35 .50 Cond. Joint Revised .0375 .0875 .8750

.10 .015 .10 .035 .70 .350 P(FInc) = .400

Chapter 19: Decision Analysis 28

Expected Value With Sample Information = 21,425.55 EVSI = EVWSI - EMV = 21,425.55 - 18,750 = 2,675.55

Chapter 19: Decision Analysis 29

19.24 Build Don't

Chosen(.20) 12,000 -1,000

Not Chosen(.80) -8,000 2,000

EMV -4,000 1,400

Decision: Based on Max EMV = Max {-4000, 1400} = 1,400, choose "Don't Build" as a strategy. Forecast Chosen: State Chosen Not Chosen Prior .20 .80 Cond. .45 Joint .090 Revised .2195 .7805

.40 .320 P(FC) = .410

Forecast Not Chosen: State Chosen Not Chosen Prior .20 .80 Cond. .55 Joint .110 Revised .1864 .8136

.60 .480 P(FC) = .590

Chapter 19: Decision Analysis 30

Expected Value With Sample Information = 1,400.09 EVSI = EVWSI - EMV = 1,400.09 - 1,400 = .09

Вам также может понравиться