Chapter 18 Decision Making Under Uncertainty CHAPTER OVERVIEW AND OBJECTIVES This chapter provides an introduction to decision strategies including structuring a decision problem into various actions under consideration and states of nature describing the uncertain future. Defining a decision table by using utility values, rather than dollars, is presented. The concept of decision trees and using posterior (revised) probabilities in a decision strategy is discussed and illustrated using several examples. Having completed this material, the student should be able to: 1. Discuss and implement various decision strategies including the minimax, maximax and expected payoff procedures. 2. Evaluate the risk associated with a given strategy. 3. Conduct a sensitivity analysis for a decision problem. 4. Construct and apply utility curves. 5. Use a decision tree to structure a decision application and revise prior probabilities for each state of nature. 6. Discuss and determine the expected value of perfect information (EVPI) and the expected value of sample information (EVSI). 710 Instructor's Manual Chapter 18 Glossary actions (alternatives). The events in a decision problem over which the decision maker does have control, such as "purchase" or "lease". admissable action. An action for which no other action under consideration dominates it. chance node. A point in a decision tree that is not under the control of the decision maker. Leading away from a chance node are states of nature, not possible actions. decision node. A point in a decision tree that is under the control of the decision maker. Leading away from a decision node are possible actions, not states of nature. decision tree. A method of representing a multistage decision problem, containing decision nodes and/or chance nodes connected by straight lines. dominated action. Action Aj is dominated by action Ai if the payoff for Aj is less than or equal to that for Ai under each state of nature. efficiency of sample information. The ratio EVSI/EVPI times 100. expected value of perfect information (EVPI). The amount a decision maker would be willing to pay for a perfect predictor. expected value of sample information (EVSI). The expected payoff when obtaining sample information minus the expected payoff without obtaining sample information. maximax strategy. A decision strategy that selects the action having the largest possible payoff. 710 Chapter 18 minimax strategy. 711 A decision strategy (usually conservative) that minimizes the maximum possible opportunity loss. opportunity loss (Lij). The difference between the payoff for action i and the payoff for the action that would have the largest payoff under state of nature j. payoff. The profit associated with taking a particular action given that a specific state of nature has occurred. posterior probability. A revised probability based upon additional information. prior probability. risk. An initial estimate for the probability of an event. The variance of the possible payoffs for a particular action. risk avoider. A decision maker who prefers a smaller expected payoff with a small risk over a larger expected payoff with a large risk. risk neutral. A decision maker who maximizes expected utility by maximizing the expected payoff. risk taker. A decision maker who prefers an action with a possible large payoff even if that action has a large risk. states of nature. The uncertain events over which the decision maker has no control. utility value. For a particular outcome, a measure of the attractiveness and the risk associated with the corresponding dollar amount. 711 712 Instructor's Manual 18.1 States of Nature S1 S2 S3 S4 A1 500 500 500 500 A2 400 500 10000 20000 A3 -500 0 10000 30000 Under state S1, action A1 is the best. or A2 is the best. Under state S2, action A1 Under state S3, action A2 or A3 is the best. Under state S4, action A3 is the best. 18.2 a) Uncertainty b) Certainty c) Certainty d) Uncertainty 18.3 States of Nature Action 7 8 9 10 11 12 5 5 = 5 = 5 = 5 = 5 = 5 = x 25 145 -20 x 25 155 -30 x 25 165 -40 x 25 175 -50 x 25 185 -60 x 25 195 -70 6 6 = 6 = 6 = 6 = 6 = 6 = x 25 145 5 x 25 155 -5 x 25 165 -15 x 25 175 -25 x 25 185 -35 x 25 195 -45 7 8 9 10 11 12 13 30 30 30 30 30 30 30 20 45 45 45 45 45 45 10 35 60 60 60 60 60 0 25 50 75 75 75 75 -10 15 40 65 90 90 90 -20 5 30 55 80 105 105 712 13 5 x 25 - 205 = -80 6 x 25 - 205 = -55 -30 -5 20 45 Chapter 18 713 70 95 120 18.4 States of Nature Action A1 A2 A3 S1 88,000 - 700 = 87,300 85,000 - 700 = 84,300 82,000 - 700 = 81,300 S2 S3 88,000 - 1,400 = 86,600 85,000 - 1,400 = 83,600 82,000 - 1,400 = 80,600 S4 88,000 - 2,100 = 85,900 85,000 - 2,100 = 82,900 82,000 - 2,100 = 79,900 S5 88,000 - 2,800 = 85,200 85,000 - 2,800 = 82,200 82,000 - 2,800 = 79,200 88,000 - 3,500 = 84,500 85,000 - 3,500 = 81,500 82,000 - 3,500 = 78,500 S6 88,000 - 4,200 = 83,800 85,000 - 4,200 = 80,800 82,000 - 4,200 = 77,800 18.5 States of Nature Action S1(100) S2(125) S3(150) A1(100) .90 x 100 = 90 .90 x 100 = 90 .90 x 100 = 90 A2(125) .90 x 100 1.10 x 25 .90 x 100 1.10 x 50 .90 x 125 = 112.5 .90 x 125 1.10 x 25 = 85 .90 x 125 = 112.5 .90 x 150 = 135 A3(150) 18.6 = 62.5 = 35 Opportunity loss table: States of Nature S1 S2 S3 S4 A1 0 0 2 6 A2 1 1 1 2 A3 1 2 0 1 A4 3 3 1 0 713 714 Instructor's Manual Action Maximum Opportunity Loss A1 6 A2 2 A3 2 A4 3 The minimax decision is either action A2 or A3. 18.7 a) No, because under S2 there is no 0. 18.8 b) No, because there is a negative value in the table. Payoff table: States of Nature Action Complete Loss No Loss Insure -1,000 -1,000 Not Insure -200,000 0 Opportunity loss table: Action S1 S2 A1 0 1,000 A2 199,000 0 The minimax decision is A1. 18.9 The maximax decision is to choose the action having the largest payoff. Therefore, a high inventory level would be the maximax decision. Opportunity loss table: States of Nature Action S1 S2 S3 A1 0 3000 7000 A2 3000 0 4000 714 A3 8000 Chapter 18 0 4000 Action Maximum Opportunity Loss A1 7000 A2 4000 A3 8000 The minimax decision is A2. 18.10 The maximax decision is A3 (order 150 gallons a week). Opportunity loss table: States of Nature Action S1 S2 S3 A1 0 22.5 45 A2 27.5 A3 55 0 22.5 27.5 0 Action Maximum Opportunity Loss A1 45 A2 27.5 A3 55 The minimax decision is A2 (order 125 gallons a week). 18.11 P(S1) = .2 P(S2) = .4 P(S3) = .4 The expected payoffs for the 3 actions are: Expected Payoff A1: (.2)(40) + (.4)(8) + (.4)(0) = A2: (.2)(10) + (.4)(60) + (.4)(20) = A3: (.2)(0) + (.4)(20) + (.4)(80) 11.2 34 = 40.0 Risk 715 715 716 Instructor's Manual A1: (.2)(40)2 + (.4)(8)2 + (.4)(0)2 - (11.2)2 = 220.16 A2: (.2)(10)2 + (.4)(60)2 + (.4)(20)2 - (34)2 = 464.00 A3: (.2)(0)2 + (.4)(20)2 + (.4)(80)2 - (40)2 = 1120.00 18.12 Opportunity loss table: States of Nature a) S1 S2 S3 A1 0 530 100 A2 1500 0 250 A3 2500 400 0 A4 2700 700 220 Action Maximum Opportunity Loss A1 530* A2 1500 A3 2500 A4 2700 * The minimax decision is A1 b) E(A1) = 5500(.25) + 2670(.50) + 1300(.25) = 3035 E(A2) = 4000(.25) + 3200(.50) + 1150(.25) = 2887.5 E(A3) = 3000(.25) + 2800(.50) + 1400(.25) = 2500 E(A4) = 2800(.25) + 2500(.50) + 1180(.25) = 2245 The decision based on maximum expected payoff is A1. c) Risk (A1) = 55002(.25) + 26702(.50) + 13002(.25) - 30352 = 2,338,225 Risk (A2) = 40002(.25) + 32002(.50) + 11502(.25) - 2887.52 = 1,112,968.75 Risk (A3) = 30002(.25) + 28002(.50) + 14002(.25) - 25002 716 Chapter 18 = 410,000 Risk (A4) = 28002(.25) + 25002(.50) + 11802(.25) - 22452 = 393,075 d) The decision is A4 based on minimum risk. 18.13 Opportunity loss table: States of Nature a) Action S1 S2 S3 A1 150 60 10 A2 100 20 0 A3 50 0 35 A4 0 40 160 Action Maximum Opportunity Loss A1 150 A2 100 A3 50* A4 160 * The minimax decision is A3. b) E(A1) = 100(.35) + 100(.50) + 100(.15) = 100 E(A2) = 150(.35) + 140(.50) + 110(.15) = 139 E(A3) = 200(.35) + 160(.50) + 75(.15) = 161.25 E(A4) = 250(.35) + 120(.50) + (-50)(.15) = 140 The decision is A3 based on the maximum expected payoff. c) Risk (A1) = 0 Risk (A2) = 1502(.35) + 1402(.50) + 1102(.15) - 1392 = 169 Risk (A3) = 2002(.35) + 1602(.50) + 752(.15) - 161.252 = 1642.1875 717 717 718 Instructor's Manual Risk (A4) = 2502(.35) + 1202(.50) + 502(.15) - 1402 = 9850 18.14 A1 = low price A2 = high price S1 = 1 year P(S1) = .4 S2 = 2 years P(S2) = .4 S3 = 3 years P(S3) = .2 The expected payoffs for the 3 actions are: A1: (.4)(10,000) + (.4)(15,000) + (.2)(20,000) = 14,000 A2: (.4)(3,000) + (.4)(10,000) + (.2)(30,000) = 11,200 A2 is the maximum expected payoff decision. 18.15 A1 = low investment A2 = medium investment A3 = high investment The expected payoffs for the 3 actions are: A1: (.2)(300,000) + (.5)(400,000) + (.3)(500,000) = 410,000 A2: (.2)(-100,000) + (.5)(900,000) + (.3)(1,000,000) = 730,000 A3: (.2)(-400,000) + (.5)(300,000) + (.3)(3,000,000) = 970,000 The maximum expected payoff action is A3. 18.16 E(insure) = .05(-1000) + .95(-1000) = -1000 E(not insure) = .05(-200,000) + .95(0) = -10,000 The decision is to buy the insurance. Risk (insure) = .05(-1000)2 + .95(-1000)2 - (-1000)2 = 0 Risk (not insure) = .05(-200,000)2 + .95(0) - (-10,000)2 = 1,900,000,000 18.17 Expected payoff using a perfect predictor is: (.2)(500) + (.4)(400) + (.3)(150) + (.1)(200) = 325 EVPI = 325 - 180 (maximum expected payoff) = 145 718 Chapter 18 18.18 The expected payoffs for the 3 actions are: Expected Payoff A1: (.3)(30) + (.4)(2) + (.3)(2) = 10.4 A2: (.3)(5) + (.4)(10) + (.3)(1) = 5.8 A3: (.3)(1) + (.4)(1) + (.3)(5) 2.2 = A1 is the maximum expected payoff decision. States of Nature Maximum Payoff P(Si) S1 30 .3 S2 10 .4 S3 5 .3 The expected payoff using a perfect predictor is: (.3)(30) + (.4)(10) + (.3)(5) = 14.5 EVPI = 14.5 - 10.4 = 4.1 18.19 Payoff Table: States of Nature Action 1500 2000 2500 1500 750 750 750 2000 750 - .3(500) = 600 750 - .3(1000) = 450 1000 1000 1000 - .3(500) = 850 1250 2500 .40 x 750 + .40 x 1000 + .20 x 1250 = 950 E(A1) = .4(750) + .4(750) + .2(750) = 750 E(A2) = .4(600) + .4(1000) + .2(1000) = 840 E(A3) = .4(450) + .4(850) + .2(1250) = 770 EVPI = 950 - 840 = 110 719 719 720 Instructor's Manual $110 is the maximum amount that the manager would be willing to pay for perfect information. 18.20 The maximum payoff using a perfect predictor = 40,000 + 25,000 = 65,000 18.21 P(S1) = .30 P(S2) = .40 P(S3) = .30 The expected payoffs for the 3 actions are: Expected Payoff A1: (.3)(10,500) + (.4)(8,000) + (.3)(5,000) = 7,850 A2: (.3)(-1,000) + (.4)(15,000) + (.3)(10,000) = 8,700 A3: (.3)(-4,000) + (.4)(3,000) + (.3)(30,000) = 9,000 The maximum expected payoff decision is given by A3. States of Nature Maximum Payoff P(Si) S1 10,500 .3 S2 15,000 .4 S3 30,000 .3 The expected payoff using a perfect predictor is: (.3)(10,500) + (.4)(15,000) + (.3)(30,000) = 18,150 EVPI = 18,150 - 9,000 = 9,150 An upper limit for the price of advice from a consultant would be the EVPI. 18.22 5000 x .25 + 3200 x .5 + 2000 x .25 = 3350 E(A1) = 3000 E(A2) = 3000 E(A3) = 2550 E(A4) = 2500 EVPI = 3350 - 3000 = 350 18.23 250 x .35 + 160 x .5 + 110 x .15 = 184 E(A1) = 100 E(A2) = 139 E(A3) = 161.25 720 E(A4) = 140 Chapter 18 721 EVPI = 184 - 161.25 = 22.75 A1 is inadmissible because A2 dominates A1. 18.24 a) U(x) is a utility function for a risk taker. Note the U(X) lies below the straight line connecting (0, 0) and (1, 100). b) U(x) is a utility function for a risk neutral decision maker. c) U(x) is not a utility function since it is a decresing function of x. 18.25 a) The expected payoffs for the 3 actions are: Expected Payoff A1: (.25)(4 + 100 + 49 + 9) = 40.5 A2: (.25)(81 + 25 + 36 + 25) = 41.75 A3: (.25)(100 + 16 + 25 + 9) = 37.5 A2 is the decision based on the maximum expected payoff. b) Table of the utility of the payoff: States of Nature Action S1 S2 S3 S4 A1 20 100 70 30 A2 90 50 60 50 A3 100 40 50 30 The expected utility of the payoff for the 3 actions are: Expected Utility of Payoff A1: (.25)(20 + 100 + 70 + 30) = 55 A2: (.25)(90 + 50 + 60 + 50) = 62.5 A3: (.25)(100 + 40 + 50 + 30) = 55 A2 is the decision based on the maximum expected utility. 721 722 Instructor's Manual 18.26 E(A1) = (.2)(50) + (.4)(10) + (.3)(30) + (.1)(10) = 24 E(A2) = (.2)(20) + (.4)(20) + (.3)(30) + (.1)(60) = 27 E(A3) = (.2)(10) + (.4)(50) + (.3)(10) + (.1)(20) = 27 The decision is A2 or A3 based on the maximum expected payoff. Table of utility values: States of Nature S1 S2 S3 S4 A1 85 15 45 15 A2 3 3 45 0 A3 15 85 15 3 E(A1) = (.2)(75) + (.4)(15) + (.3)(45) + (.1)(15) = 36 E(A2) = (.2)(3) + (.4)(3) + (.3)(45) + (.1)(9) = 40.5 E(A3) = (.2)(15) + (.4)(75) + (.3)(15) + (.1)(3) = 40.5 The decision is A2 or A3 based on the maximum expected utility of the payoff. The decision based on the maximum expected payoff is equivalent to the decision based on the maximum expected utility of the payoff. 18.27 Table of utility values: States of Nature Action S1 S2 S3 S4 A1 60 12 36 12 A2 24 24 36 72 A3 12 60 12 24 E(A1) = (.2)(60) + (.4)(12) + (.3)(36) + (.1)(12) = 28.8 E(A2) = (.2)(24) + (.4)(24) + (.3)(36) + (.1)(12) = 32.4 722 Chapter 18 723 E(A3) = (.2)(12) + (.4)(60) + (.3)(12) + (.1)(24) = 32.4 The decision is A2 or A3 based on the maximum expected utility of the payoff. The decision based on the maximum expected utility does not differ from that obtained in Exercise 18.26. 18.28 (1 - P) • U(0) + P • U(500,000) = U(50,000) (1 - P) • 0 + P • (100) = 75 P = .75 18.29 Table of utility values: States of Nature S1 S2 S3 A1 3.699 3.477 3 A2 3.643 3.505 3.079 A3 3.505 3.447 3.146 A4 3.477 3.398 3.301 E(A1) = .25(3.699) + .50(3.477) + .25(3) = 3.413 E(A2) = .25(3.643) + .50(3.505) + .25(3.079) = 3.433 E(A3) = .25(3.505) + .50(3.447) + .25(3.146) = 3.386 E(A4) = .25(3.477) + .50(3.398) + .25(3.30) = 3.394 a) The decision is A2 based on the maximum expected utility of the payoff. b) The manager is a risk avoider. 723 724 Instructor's Manual 18.30 a) The expected payoffs for the 3 actions are: Expected Payoff A1: (.25)(151) + (.30)(33) + (.40)(95) + (.05)(40) = 87.65 A2: (.25)(75) + (.30)(75) + (.40)(97) + (.05)(180) = 89.05 A3: (.25)(29) + (.30)(162) + (.40)(30) + (.05)(50) = 70.35 b) The decision based on the maximum expected payoff is A2. Table of the utility of the payoff: States of Nature Action S1 S2 S3 S4 A1 54.98 05.62 27.44 07.50 A2 19.25 19.25 28.31 71.55 A3 04.63 61.09 04.87 10.48 The expected utility payoff for the 3 actions are: Expected Utility Payoff A1: (.25)(54.98) + (.30)(05.62) + (.40)(27.44) + (.05)(07.50) = 26.78 A2: (.25)(19.25) + (.30)(19.25) + (.40)(28.31) + (.05)(71.55) = 25.49 A3: (.25)(04.63) + (.30)(61.09) + (.40)(04.87) + (.05)(10.48) = 21.96 The decision based on the maximum expected utility of the payoff is A1. 724 Chapter 18 18.31 18.32 Sum of paths = (.2)(.5) + (.1)(.3) + (.1)(.5) + (.3)(.2) + (.3)(.1) = .27 P(AiB) P(A1B) P(A2B) P(A3B) P(A4B) P(A5B) = = = = = = ith path/sum (.2)(.5)/.27 (.1)(.3)/.27 (.1)(.5)/.27 (.3)(.2)/.27 (.5)(.1)/.27 of paths = .371 = .111 = .185 = .222 = .111 725 725 726 Instructor's Manual 18.33 18.34 726 Chapter 18 18.35 Expected payoffs for the 4 actions are: Expected Payoff A1: (.15)(1451) + (.25)(1840) + (.25)(2050) + (.35)(2300) = 1995.15 A2: (.15)(-1091) + (.25)(1685) + (.25)(2430) + (.35)(2900) = 1880.1 A3: (.15)(-2015) + (.25)(1100) + (.25)(3060) + (.35)(3561) = 1984.1 A4: (.15)(-3460) + (.25)(-1350) + (.25)(3340) + (.35)(4300) = 1483.5 18.36 a) 727 727 728 Instructor's Manual P(I1) = sum of the branches P(I2) = sum of the branches = .23 = .31 P(S1I1) = .12/.23 = .522 P(S1I2) = .015/.31 = .048 P(S2I1) = .025/.23 = .109 P(S2I2) = .175/.31 = .565 P(S3I1) = .05/.23 = .217 P(S3I2) = .05/.31 = .161 P(S4I1) = .035/.23 = .152 P(S4I2) = .07/.31 = .226 P(I3) = sum of the branches = .2625 P(I4) = sum of the branches = .1975 P(S1I3) = .0075/.2625 = .029 P(S1I4) = .0075/.1975 = .038 P(S2I3) = .025/.2625 = .095 P(S2I4) = .025/.1975 = .127 P(S3I3) = .125/.2625 = .476 P(S4I3) = .105/.2625 = .4 P(S3I4) = .025/.1975 = .127 P(S4I4) = .14/.1975 = .709 728 Chapter 18 729 729 730 Instructor's Manual Consultant's fee = $400 EVSI = 2445.27 - 400 - 1995.15 = 50.12 b) States of Nature Maximum Payoff P(Si) S1 1451 .15 S2 1840 .25 S3 3340 .25 S4 4300 .35 Average payoff using a perfect predictor: (1451)(.15) + (1840)(.25) + (3340)(.25) + (4300)(.35) = 3017.65 EVPI = 3017.65 - 1995.15 = 1022.5 The efficiency of the sample information: (EVSI/EVPI)(100) = (50.12/1022.5)(100) = 4.9% c) The consultant's service would be slightly better than using the action which gives the largest expected payoff without sample information. The low efficiency indicates that perhaps another consultant would be more profitable. 18.37 P(B) = (.3)(.1) + (.1)(.3) + (.2)(.3) + (.5)(.1) + (.5)(.2) = .03 + .03 + .06 + .05 + .10 = .27 P(A1B) = P(BA1) P(A1) / P(B) = .03/.27 = .111 P(A2B) = P(BA2) P(A2) / P(B) = .03/.27 = .111 P(A3B) = P(BA3) P(A3) / P(B) = .06/.27 = .222 P(A4B) = P(BA4) P(A4) / P(B) = .05/.27 = .185 P(A5B) = P(BA5) P(A5) / P(B) = .10/.27 = .370 730 Chapter 18 18.38 P[ProfitMkt] = .6(.7)/[.6(.7) + .3(.3)] = .42/.51 = .824 The probability that the stores market a particular fashion is .51. 18.39 P(life) = .4(.05) + .33(.08) + .27(.10) = .0734 P(Hlife) = .33(.08)/.0734 = .0264/.0734 = .3597 18.40 States of Nature Maximum Payoff P(Si) S1 44 .1 S2 27 .3 S3 40 .4 S4 45 .2 The expected payoff using a perfect predictor is: (.1)(44) + (.3)(27) + (.4)(40) + (.2)(45) = 37.5 EVPI = (expected payoff using a perfect predictor) - (expected payoff) = 37.5 - 32.4 = 5.1 EVSI = 1.5 Efficiency of the sample information = (EVSI/EVPI)100 = (1.5/5.1)100 = 29.4% 18.41 P(B) = .1(.6) + .3(.2) + .4(.3) + .2(.1) = .26 P(S1B) = (.1)(.6)/.26 = .2308 P(S2B) = (.3)(.2)/.26 = .2308 P(S3B) = (.4)(.3)/.26 = .4615 P(S4B) = (.2)(.1)/.26 = .0769 731 731 732 Instructor's Manual Choose action A2. 18.42 P(G) = (.10)(.04) + (.20)(.06) + (.20)(.08) + (.40)(.03) + (.10)(.09) = .004 + .012 + .016 + .012 + .009 = .053 where G = the event that an error is found on a legal document. P(AG) = .004/.053 = .075 P(BG) = .012/.053 = .226 P(CG) = .016/.053 = .302 P(DG) = .012/.053 = .226 P(EG) = .009/.053 = .170 Secretary C would have the highest probability of having typed the error. 18.43 (1/3)(15000 + 5000 + 1000) = 7000 EVPI = 7000 - 4000 = 3000 Since 2000 < 3000, it is worthwhile to conduct the experiment. 732 Chapter 18 18.44 P(I1) = .4(.8) + .4(.3) + .2(.3) = .32 + .12 + .06 = .50 P(I2) = .4(.1) + .4(.5) + .2(.1) = .04 + .2 + .02 = .26 P(I3) = .4(.1) + .4(.2) + .2(.6) = .04 + .08 + .12 = .24 P(S1I1) = .32/.50 = .64 P(S2I1) = .12/.50 = .24 P(S3I1) = .06/.50 = .12 P(S1I2) = .04/.26 = .15 P(S2I2) = .2/.26 = .77 P(S3I2) = .02/.26 = .08 P(S1I3) = .04/.24 = .17 P(S2I3) = .08/.24 = .33 P(S3I3) = .12/.24 = .50 EVPI is 64.651 - 62.8 = 1.851 thousand. cost of the survey, is too expensive. 733 Therefore, $2000, the 733 734 Instructor's Manual 734 Chapter 18 18.45 a) 735 The expected payoffs for the 4 actions are: Expected Payoff A1: (.3)(25) + (.2)(25) + (.5)(25) = 25 A2: (.3)(12.5) + (.2)(32.5) + (.5)(32.5) = 26.5 A3: (.3)(0) + (.2)(40) + (.5)(40) = 28 A4: (.3)(-25) + (.2)(15) + (.5)(55) = 23 The maximum expected payoff decision is given by A3. The expected payoffs when the consultant predicts I1 are: Expected Payoff A1: (.54)(25) + (.2)(25) + (.26)(25) = 25 A2: (.54)(12.5) + (.2)(32.5) + (.26)(32.5) = 21.7 A3: (.54)(0) + (.2)(40) + (.26)(40) = 18.4 A4: (.54)(-25) + (.2)(15) + (.26)(55) = 3.8 The maximum expexcted payoff decision when the consultant predicts I1 is given by A1. The expected payoffs when the consultant predicts I2 are: Expected Payoff A1: (.231)(25) + (.385)(25) + (.385)(25) = 25.025 A2: (.231)(12.5) + (.385)(32.5) + (.385)(32.5) = 27.913 A3: (.231)(0) + (.385)(40) + (.385)(40) = 30.8 A4: (.231)(-25) + (.385)(15) + (.385)(55) = 21.175 The maximum expected payoff decision when the consultant predicts I2 is given by A3. The expected payoffs when the consultant predicts I3 are: Expected Payoff A1: (.086)(25) + (.057)(25) + (.857)(25) 735 = 25 736 Instructor's Manual A2: (.086)(12.5) + (.057)(32.5) + (.857)(32.5) = 30.78 A3: (.086)(0) + (.057)(40) + (.857)(40) = 36.56 A4: (.086)(-25) + (.057)(15) + (.857)(55) = 45.84 The maximum expected payoff decision when the consultant predicts I3 is given by A4. Net expected gain of hiring the consultant (EVSI) is: (.39)(25) + (.26)(30.8) + (.35)(46.2) - 2.5 - 28 = 3.428 b) Given a perfect predictor, the following payoffs are possible: Si Maximum Payoff P(Si) S1 25 .3 S2 40 .2 S3 55 .5 The expected payoff using a perfect predictor is: (.3)(25) + (.2)(40) + (.5)(55) = 43 EVPI = 43 - 28 = 15 Efficiency of the sample information = (EVSI/EPVI)(100) = (3.428/15)(100) = 22.85% 18.46 The expected payoffs for the 3 actions are: A1: A2: A3: The Expected Payoff (.4)(10,000) + (.3)(13,000) + (.3)(16,000) = 12,700 (.4)(8,000) + (.3)(23,000) + (.3)(25,000) = 17,600 (.4)(8,000) + (.3)(20,000) + (.3)(40,000) = 21,200 decision based on the maxium e expected payoff is A3. 736 Chapter 18 P(I1) = sum of the branches = .5 P(S1I1) = .32/.5 = .64 P(S1I2) = .04/.22 = .182 P(S2I1) = .12/.5 = .24 P(S2I2) = .12/.22 = .545 P(S3I1) = .06/.5 = .12 P(S3I2) = .06/.22 = .273 737 737 738 Instructor's Manual P(I2) = sum of the branches = .22 P(S1I2) = .04/.22 = .182 P(S2I2) = .12/.22 = .545 P(S3I2) = .06/.22 = .273 738 Chapter 18 P(I3) = sum of the branches = .28 P(S1I3) = .04/.28 = .143 P(S2I3) = .06/.28 = .214 P(S3I3) = .18/.28 = .643 739 739 740 Instructor's Manual I2 .22 I3 .28 740 Chapter 18 741 Note that A3 gives the largest expected payoff for each of the predictions I1, I2, and I3 of the consultant. The decision based on the maximum expected payoff is A3. Therefore, it would not be worth paying $2,000 for the financial planner's advice. 18.47 a) Opportunity loss table: States of Nature b) S1 S2 S3 S4 A1 0 0 10 20 A2 4 2 0 10 A3 4 4 0 0 A4 6 6 14 5 Action Maximum Opportunity Loss A1 20 A2 10 A3 4 A4 14 The minimax decision is action A3. c) Action Maximum Payoff A1 40 A2 50 A3 60 A4 55 The maximax decision is A3. 741 742 Instructor's Manual 18.48 Payoff table: S1 (5,000) -100,000 S2 (10,000) 0 S3 (15,000) 100,000 S4 (20,000) 200,000 S5 (25,000) 300,000 200,000 200,000 200,000 200,000 200,000 S1 S2 S3 S4 S5 A1 300,000 200,000 100,000 0 0 A2 0 0 0 0 100,000 A1 (introduce software) A2 (don’t introduce software) Opportunity loss table: Action Maximum Opportunity Loss A1 300,000 A2 100,000 The minimax decision is A2. 18.49 i) With the minimax strategy, the change of probabilities will not change the decision. ii) Decision based on a maximum expected payoff. a) A1: (.1)(250) + (.5)(320) + (.2)(350) + (.2)(400) = 335 A2: (.1)(150) + (.5)(260) + (.2)(300) + (.2)(370) = 279 A3: (.1)(120) + (.5)(290) + (.2)(380) + (.2)(450) = 323 A4: (.1)(80) + (.5)(280) + (.2)(410) + (.2)(550) b) = 340 A1: (.1)(250) + (.5)(320) + (.1)(350) + (.3)(400) = 340 A2: (.1)(150) + (.5)(260) + (.1)(300) + (.3)(370) = 286 A3: (.1)(120) + (.5)(290) + (.1)(380) + (.3)(450) = 330 A4: (.1)(80) + (.5)(280) + (.1)(410) + (.3)(550) 742 = 354 Chapter 18 c) 743 A1: (.1)(250) + (.5)(320) + (.3)(350) + (.1)(400) = 330 A2: (.1)(150) + (.5)(260) + (.3)(300) + (.1)(370) = 272 A3: (.1)(120) + (.5)(290) + (.3)(380) + (.1)(450) = 316 A4: (.1)(80) + (.5)(280) + (.3)(410) + (.1)(550) d) = 326 A1: (.1)(250) + (.5)(320) + (.2)(350) + (.2)(400) = 335 A2: (.1)(150) + (.5)(260) + (.2)(300) + (.2)(370) = 279 A3: (.1)(120) + (.5)(290) + (.2)(380) + (.2)(450) = 323 A4: (.1)(80) + (.5)(280) + (.2)(410) + (.2)(550) e) = 342 A1: (.2)(250) + (.5)(320) + (.2)(350) + (.1)(400) = 320 A2: (.2)(150) + (.5)(260) + (.2)(300) + (.1)(370) = 257 A3: (.2)(120) + (.5)(290) + (.2)(380) + (.1)(450) = 290 A4: (.2)(80) + (.5)(280) + (.2)(410) + (.1)(550) = 295 Since three out of five sets of probabilities make the same decision, it is not extremely sensitive to changing the probabilities. Action A4 is the optimal action for sets 1, 2, and 4. Action A1 is the optimal action for sets 3 and 5. Summary of Sensitivity Analysis Expected Payoff P(S1) P(S2) P(S3) P(S4) A1 A2 A3 A4 .1 .5 .2 .2 335 279 323 340* .1 .5 .1 .3 340 286 330 354* .1 .5 .3 .1 330* 272 316 326 .1 .5 .2 .2 335 279 323 342* .1 .5 .2 .1 320* 257 290 295 743 744 Instructor's Manual 18.50 a) b) All the actions are admissible. Opportunity loss table: States of Nature Action S1 S2 A1 1020 375 0 A2 620 0 90 A3 0 215 380 Action S3 Maximum Opportunity Loss A1 1020 A2 620 A3 380* * The minimax decision is A3. c) E(A1) = .3(830) + .55(750) + .15(710) = 768 E(A2) = .3(1230) + .55(1125) + .15(620) = 1080.75 E(A3) = .3(1850) + .55(910) + .15(330) = 1105 The decision is A3 based on the maximum expected payoff. d) Risk (A1) = .3(830)2 + .55(750)2 + .15(710)2 - 7682 = 1836 Risk (A2) = .3(1230)2 + .55(1125)2 + .15(620)2 - 1080.752 = 39603.1875 Risk (A3) = .3(1850)2 + .55(910)2 + .15(330)2 - 11052 = 277515 e) The decision is A1 based on minimum risk. f) .3(1850) + .55(1125) + .15(710) = 1280.25 EVPI = 1280.25 - 1105 = 175.25 744 Chapter 18 745 18.51 One example is: 0 for x 0 U(x) .036x for 0 x 1000 36 64(.001x 1)2 for 1000 x 2000 This utility function is scaled so that its values are between 0 and 100. 18.52 P(S1) = .2 P(S2) = .2 P(S3) = .4 P(S4) = .2 The expected payoffs for the 4 actions are: A1: (.2)(9) + (.2)(19) + (.4)(29) + (.2)(39) = 25 A2: (.2)(9) + (.2)(24) + (.4)(39) + (.2)(54) = 33 A3: (.2)(5) + (.2)(25) + (.4)(45) + (.2)(65) = 37 Risk A1: (.2)(9)2 + (.2)(19)2 + (.4)(29)2 + (.2)(39)2 - 252 = 104 A2: (.2)(9)2 + (.2)(24)2 + (.4)(39)2 + (.2)(54)2 - 332 = 234 A3: (.2)(5)2 + (.2)(25)2 + (.4)(45)2 + (.2)(65)2 - 372 = 416 18.53 EVSI = 21 - 18 = 3 EVPI = 30 - 18 = 12 Efficiency of the sample information = (EVSI/EVPI)(100) = 95.45% 18.54 P(I1) = .3(.6) + .55(.2) + .15(.2) = .18 + .11 + .03 = .32 P(S1 I1) = .18/.32 = .56 P(S2I1) = .11/.32 = .34 P(S3I1) = .03/.32 = .09 745 746 Instructor's Manual Under I1: E(A1) = .56(830) + .34(750) + .10(710) = 790.8 E(A2) = .56(1230) + .34(1125) + .10(620) = 1133.3 E(A3) = .56(1850) + .34(910) + .10(330) = 1378.4 The decision is A3 under the revised probabilities. 18.55 P(I2) = .3 x .1 + .55 x .5 + .15 x .3 = .03 + .275 + .045 = .35 P(I3) = .3 x .3 + .55 x .3 + .15 x .5 = .09 + .165 + .075 = .33 P(S1I2) = .03/.35 = .08 P(S1I3) = .09/.33 = .27 P(S2I2) = .275/.35 = .79 P(S2I3) = .165/.33 = .5 P(S3I2) = .045/.35 = .13 P(S3I3) = .075/.33 = .23 Under I2: E(A1) = .08(830) + .79(750) + .13(710) = 751.2 E(A2) = .08(1230) + .79(1125) + .13(620) = 1067.75 E(A3) = .08(1850) + .79(910) + .13(330) = 909.8 Under I3: E(A1) = .27(830) + .5(750) + .23(710) = 762.4 E(A2) = .27(1230) + .5(1125) + .23(620) = 1037.2 E(A3) = .27(1850) + .5(910) + .23(330) = 1030.4 The maximum payoff with consultant = .32(1378.4) + .35(1067.75) + .33(1037.2) = 1157.0765 The expected payoff without the consultant is 1105 (from A3). Since 1157.0765 - 1105 = 52.0765 is less than 350, it is not worthwhile to have the consultant's service. 18.56 Let E represent the event that an electronic component is defective. 746 Chapter 18 747 P(E) = (.009)(.48) + (.005)(.22) + (.002)(.30) = .00432 + .0011 + .0006 = .006 P(AE) = .00432/.006 = .72 P(BE) = .0011/.006 = .18 P(CE) = .0006/.006 = .10 18.57 a) Risk taker b) Risk neutral c) Risk avoider 18.58 a) Opportunity loss table: States of Nature S1 S2 S3 A1 350 100 0 A2 100 0 150 A3 0 600 650 Action Maximum Opportunity Loss A1 350 A2 150 A3 650 The minimax decision is A2. b) E(A1) = .40(450) + .35(400) + .25(350) = 407.50 E(A2) = .40(700) + .35(500) + .25(200) = 505.00 E(A3) = .40(800) + .35(-100) + .25(-300) = 210.00 A2 would be the decision based on the maximum expected payoff. c) Risk (A1) = (450)2(.40) + (400)2(.35) + (350)2(.25) - (407.50)2 = 1568.75 747 748 Instructor's Manual Risk (A2) = (700)2(.40) + (500)2(.35) + (200)2(.25) - (505.00)2 = 38475.00 Risk (A3) = (800)2(.40) + (-100)2(.35) + (-300)2(.25) - (210.00)2 = 237900.00 18.59 Let A represent drives own car. Let B represent uses car pool. Let C represent uses city bus. Let M represent male. P[M] = P[MA]•P[A] + P[MB]•P[B] + P[MC]•P[C] = (.70)(.55) + (.30)(.25) + (.20)(.20) = .50 P[AM] = P[MA] P[A] / P(M) = .385/.50 = .77 748
© Copyright 2025 Paperzz