# Making choices Dr. Yan Liu

## Presentation on theme: "Making choices Dr. Yan Liu"— Presentation transcript:

Making choices Dr. Yan Liu
Department of Biomedical, Industrial & Human Factors Engineering Wright State University

Expected Monetary Value (EMV)
One way to choose among risky alternatives is to pick the alternative with the highest expected value (EV). When the objective is measured in monetary values, the expected money value (EMV) is used EV is the mean of a random variable that has a probability distribution function (Discrete Variable) (Continuous Variable)

A2 A1 O1 C1 O2 O3 O4 C2 C3 C4 (p1) Payoff (1-p1) (p2) (1-p2) EMV(A1)=C1•p1 +C2•(1-p1) EMV(A2)=C3•p2 +C4• (1-p2)

Solving Decision Trees
Decision Trees are Solved by “Rolling Back” the Trees Start at the endpoints of the branches on the far right-hand side and move to left When encountering a chance node, calculate its EV and replace the node with the EV When encountering a decision node, choose the branch with the highest EV Continue with the same procedures until a preferred alternative is selected for each decision node

Lottery Ticket Example
You have a ticket which will let you participate in a lottery that will pay off \$10 with a 45% chance and nothing with a 55% chance. Your friend has a ticket to a different lottery that has a 20% chance of paying \$25 and an 80% chance of paying nothing. Your friend has offered to let you have his ticket if you will give him your ticket plus one dollar. Should you agree to trade? Win \$24 EMV=\$4 \$25 (0.2) Trade Ticket EMV(Trade Ticket)=24•0.2+ (-1)•0.8=\$4 Lose Ticket Result -\$1 -\$1 \$0 (0.8) Win \$10 EMV=\$4.5 \$10 (0.45) Keep Ticket EMV(Keep Ticket)=100•0.45+ (0)•0.55=\$4.5 Lose Ticket Result \$0 \$0 (0.55) Conclusion: You should keep your ticket !

Product-Switching Example
A company needs to decide whether to switch to a new product or not. The product that the company is currently making provides a fixed payoff of \$150,000. If the company switches to the new product, its payoff depends on the level of sales. It is estimated that there are about 30% chance of high-level sales (\$300,000 payoff), 50% chance of medium-level sales (\$100,000 payoff), and 20% chance of low-level sales (losing \$100,000). A survey which costs \$20,000 can be performed to provide information regarding the sales to be expected. If the survey shows high-level sales, then there are about 60% chance of high-level sales and 40% chance of medium-level sales when the company sells the product. On the other hand, if the survey shows low-level sales, then there are about 60%chance of medium-level sales and 40% chance of low-level sales when the company sells the product.

Old \$150,000 \$130,000 Survey High High \$300,000 (0.6) \$280,000 (0.5) New Medium \$100,000 (0.4) \$80,000 Old \$150,000 Perform Survey \$130,000 Survey Low Medium \$100,000 (0.6) (0.5) \$80,000 New -\$20,000 Low -\$100,000 (0.4) -\$120,000 Don’t Perform Old \$150,000 \$150,000 New High \$300,000 (0.3) \$300,000 Medium \$100,000 (0.5) \$100,000 Low -\$100,000 (0.2) -\$100,000

Old \$150,000 \$130,000 Survey High High \$300,000 (0.6) \$280,000 (0.5)
EMV= \$200,000 Survey High High \$300,000 (0.6) \$280,000 EMV= \$165,000 (0.5) New EMV(U3) =0.6•280, •80,000=\$200,000 U3 Medium \$100,000 (0.4) \$80,000 Old \$150,000 Perform Survey U1 D4 \$130,000 Survey Low EMV= \$0 Medium \$100,000 (0.6) (0.5) \$80,000 New EMV(U4) =0.6•80, •(-120,000)=\$0 -\$20,000 Low -\$100,000 (0.4) U4 -\$120,000 D1 EMV(U1) =0.5•200, •130,000=\$165,000 D2 Don’t Perform Old \$150,000 \$150,000 New High \$300,000 (0.3) \$300,000 Medium \$100,000 (0.5) \$100,000 -\$100,000 (0.2) EMV= \$120,000 U2 Low -\$100,000 EMV(U2) =0.3•300, •(100,000)+0.2•(-100,000)=\$120,000 Conclusion: Perform survey. If survey shows high-level sales, then switch the new product ; otherwise, stay with the old product

Decision Path and Strategy
Represents a possible future scenario, starting from the left-most node to the consequence at the end of a branch by selecting one alternative from a decision node and by following one outcome from a chance node. D1 U1 D2 A1 A2 O1 O2 A3 A4 Path 1 ( A1 ) A1 O1 Path 2 ( A2O1 ) Decision Paths: A3 D1 U1 Path 3 ( A2O2A3 ) A2 D2 O2 Path 4 ( A2O2A4 ) A4

Decision Path and Strategy (Cont.)
Decision Strategy The collection of decision paths connected to one branch of the immediate decision by selecting one alternative from each decision node along that path D1 U1 D2 A1 A2 O1 O2 A3 A4 Strategy 1 (A1): Decision path A1 A1 O1 Decision Strategies: Strategy 2 (A2A3): Decision paths A2O2A3, A2O1 D1 A3 U1 A2 D2 O2 A4 Strategy 3 (A2A4): Decision paths A2O2A4, A2O1

Risk Profiles Problems with Expected Value (EV) What is Risk Profile
EV does not indicate all the possible consequences The statistical interpretation of EV as the average amount obtained by “playing the game” a large number of times is not appropriate in rare cases (e.g. hazards in nuclear power plants) What is Risk Profile A graph that shows the probabilities associated with possible consequences given a particular decision strategy Indicates the relative risk levels of strategies Steps of Deriving Risk Profiles from Decision Trees Identify the decision strategies For each strategy, collapse the decision tree by multiplying out the probabilities on sequential chance branches (Don’t confuse it with solving decision trees!) Keep track of all possible consequences Summarize the probability of occurrence for each consequence

Decision Tree of the Lottery Ticket Example
Decision strategies: Trade ticket: 2) Keep ticket: \$24(0.2), -\$1(0.8) \$10(0.45), \$0(0.55) Decision Tree of the Lottery Ticket Example Keep Ticket Trade Ticket Win \$24 (0.2) Lose -\$1 \$10 \$0 (0.8) (0.45) (0.55) Pr(Payoff) Trade Ticket Keep Ticket Payoff(\$) Risk Profiles of the Lottery Ticket Example

Decision Tree of the Product-Switching Example
Old \$130,000 Survey High High (0.6) \$280,000 New (0.5) Medium (0.4) Perform Survey \$80,000 Old \$130,000 Survey Low Medium (0.6) (0.5) \$80,000 New Low (0.4) -\$120,000 Don’t Perform Old \$150,000 New High (0.3) \$300,000 Medium (0.5) \$100,000 Low (0.2) -\$100,000 Decision Tree of the Product-Switching Example 1) Don’t perform survey and keep the old product Decision Strategies: 2) Don’t perform survey and switch to the new product 3) Perform survey, and if survey is high then keep the old product 4) Perform survey, and if survey is high then switch to the new product

Don’t Perform New Medium High Low (0.3) (0.5) (0.2) -\$100,000 \$100,000
Strategy 1): Don’t perform survey and keep the old product \$150,000 (100%) Strategy 2): Don’t perform survey and switch to the new product Payoffs \$300,000 \$100,000 -\$100,000 Probabilities 0.3 0.5 0.2 Don’t Perform New Medium High Low (0.3) (0.5) (0.2) -\$100,000 \$100,000 \$300,000 Strategy 3): Perform survey and if survey high then keep the old product Perform Survey Survey High Survey Low (0.5) Old \$130,000 \$130,000 (100%) Strategy 4): Perform survey and if survey high then switch to the new product Perform Survey Survey High Survey Low (0.5) New \$130,000 Medium (0.4) \$280,000 (0.6) High \$80,000 Payoffs \$280,000 \$130,000 \$80,000 Probabilities 0.3 0.5 0.2

Risk Profiles of the Product-Switch Example
Payoff(\$) Pr(Payoff) Risk Profiles of the Product-Switch Example Strategy 1 Strategy 2 Strategy 3 Strategy 4

Cumulative Risk Profiles
A graph that shows the cumulative probabilities associated with possible consequences given a particular decision strategy Payoff(\$) Pr(Payoff≤x) Trade Ticket Keep Ticket Cumulative Risk Profiles of the Lottery Ticket Example

Dominance Deterministic Dominance
If the worst payoff of strategy B is at least as good as that of the best payoff of strategy A, then strategy B deterministically dominates strategy A May also be concluded by drawing cumulative risk profiles Pr(Payoff ≤ x) Draw a vertical line at the place where strategy B first leaves 0. If the vertical line corresponds to 100% for strategy A, then B deterministically dominates A. strategy A strategy B Payoff

Dominance (Cont.) Stochastic Dominance
If for any x, Pr(Payoff ≤ x|strategy B) ≤ Pr(Payoff ≤ x|strategy A), then B stochastically dominates A strategy A strategy B Payoff Pr(Payoff ≤ x) There is no crossing between the cumulative risk profiles of A and B, and the cumulative risk profile of B is located at the lower-right to that of A

Making Decisions with Multiple Objectives
Summer Job Example Sam has two job offers in hand. One job is to work as an assistant at a local small business. The job would pay a minimum wage (\$5.25 per hour), require 30 to 40 hours per week, and have the weekends free. The job would last for three months, but the exact amount of work and hence the amount Sam could earn were uncertain. On the other hand, he could spend weekends with friends. The other job is to work for a conservation organization. This job would require 10 weeks of hard work and 40 hours weeks at \$6.50 per hour in a national forest in a neighboring state. This job would involve extensive camping and backpacking. Members of the maintenance crew would come from a large geographic area and spend the entire 10 weeks together, including weekends. Sam has no doubts about the earnings of this job, but the nature of the crew and the leaders could make for 10 weeks of a wonderful time, 10 weeks of misery, or anything in between.

Decision Elements Objectives (and Measures) Decision to Make
Earning money (measured in \$) Having fun (measured using a constructed 5-point Likert scale; Table 4.5 at page 138) (5) Best: A large congenial group. Many new friendships made. Work is enjoyable, and time passes quickly. (4) Good: A small but congenial group of friends. The work is interesting, and time off work is spent with a few friends in enjoyable pursuits. (3) moderate: No new friends are made. Leisure hours are spent with a few friends doing typical activities. Pay is viewed as fair for the work done. (2) Bad: Work is difficult. Coworkers complain about the low pay and poor conditions. On some weekends it is possible to spend time with a few friends, but other weekends, are boring. (1) Worst: Work is extremely difficult, and working conditions are poor. Time off work is generally boring because outside activities are limited or no friends are available. Decision to Make Which job to take (In-town job or forest job) Uncertain Events Amount of fun Amount of work (# of hours per week)

Influence Diagram Overall Satisfaction Amount Fun Salary of Fun Fun
Job Decision Salary Amount of Work

Decision Tree

Analysis of the Salary Objective
EMV(Salary of Forest job) = \$2,600 EMV: EMV(Salary of In-Town job) = 0.35(2730)+0.5(2320.5)+0.15( )= \$2,422.88

Analysis of the Salary Objective
EMV(Salary of Forest job) = \$2,600 EMV: EMV(Salary of In-Town job) = 0.35(2730)+0.5(2320.5)+0.15( )= \$2,422.88 Risk Profiles: Strategies: 1) Forest Job 100% \$2,600 2) In-Town Job 35% \$2,730; 50% \$2,320.5; 15% \$2,047.5 Conclusion: For the salary objective, the forest job has higher EMV and has no risk Cumulative Risk Profiles of the Salaries

Analysis of the Fun Objective
The ratings in the original 5-point Likert scale only indicate orders of the amount of fun without carrying quantitative meanings. Therefore, the original ratings are rescaled to points to show quantitative meanings: 5(best) – 100 points, 4(Good) – 90 points, 3(Moderate) – 60 points, 2(bad) – 25 points, 1(worst) – 0 point EV: E(Fun of Forest job) =0.10(100)+0.25(90)+0.40(60)+0.20(25)+0.05(0) = 61.5 E(Fun of In-Town job) = 60

Cumulative Risk Profiles of the Fun
Analysis of the Fun Objective (Cont.) Risk Profiles: Strategies: 1) Forest Job 10% 100; 25% 90; 40% 60; 20% 30; 5% 0 2) In-Town Job 100% 60 Conclusion: For the fun objective, the forest job has higher EV but is more risky Cumulative Risk Profiles of the Fun

Sam’s dilemma: Would he prefer a slightly higher salary for sure and take a risk on how much fun the summer will be? Or otherwise, would the in-town be better, playing it safe with the amount of fun and taking a risk on how much money will be earned? Therefore, Sam needs to make a trade-off between the objectives of maximizing fun and maximizing salary.

Trade-off Analysis Combine multiple objectives into one overall objective Steps First, multiple objectives must have comparable scales Next, assign weights to these objectives (the sum of all the weights should be equal to 1) Subjective judgment Paying attention to the range of the attributes (the variables to be measured in the objectives) is crucial; Attributes having a wide range of possible values are usually important (why?) Then, calculate the weighted average of consequences as an overall score Finally, compare the alternatives using the overall score

Summer Job Example (Cont.)
Convert the salary scale to the same 0 to 100 scale used to measure fun Set \$2730 (the highest salary) = 100, and \$ (the lowest salary) =0 Then, Intermediate salary X is converted to: (X )∙100/( ) (Proportion Scoring) Assign weights to salary and fun (Ks and Kf) Sam thinks increasing salary from the lowest to the highest is 1.5 times more important than improving fun from the worst to best, hence Ks=1.5Kf , Because Ks+Kf=1  Ks=0.6, Kf=0.4

Overall Score 88.6 84.6 72.6 58.6 48.6 84.0 48.0 24.0

EV: EV(Overall Score of Forest job) =0.10(88.6)+0.25(84.6)+0.40(72.6)+0.20(58.6)+0.05(48.6) = 73.2 EV(Overall Score of In-Town job) = 0.35(84)+0.50(48)+0.15(24) = 57 Risk Profiles: The forest job stochastically dominates the in-town job Conclusion: The forest job is preferred to the in-town job Cumulative Risk Profiles of the Overall Scores

Exercise A B (0.27) A2 A1 \$8 \$0 \$15 (0.5) (0.73) (0.45) (0.55) \$4 \$10
D1 D2 A B (0.27) A2 A1 \$8 \$0 \$15 (0.5) (0.73) (0.45) (0.55) \$4 \$10 U1 U3 U2 O21 O11 O22 O12 O31 O32 1. Solve the decision tree in the figure 2. Create risk profiles and cumulative risk profiles for all possible strategies. Is one strategy stochastically dominant? Explain.

1. Solving the decision tree
A B (0.27) A2 A1 \$8 \$0 \$15 (0.5) (0.73) (0.45) (0.55) \$4 \$10 U1 U3 U2 EV(U2)=\$7.5 EV(U2)=0*0.5+15*0.5=\$7.5 EV(U1)=\$5.08 O21 O11 O22 EV(U1)=8*0.27+4*0.73=\$5.08 O12 EV(U3)=10*0.45+0*0.55=\$4.5 O31 EV(U3)=\$4.5 O32 In conclusion, according to the EV, we should choose A, and if O11 occurs, then choose A1

2. Risk Profiles and Cumulative Risk Profiles
Decision Strategies: Strategy 1: A - A1 A1 \$8 (0.27) D2 \$4 (0.73) \$8 (0.27) U1 A (0.73) \$4 D1 (0.5) Strategy 2: A – A2 (0.27) D2 \$0 \$0 (0.135) \$4 (0.73) \$15 (0.135) A2 U2 (0.5) U1 \$15 A (0.73) \$4 D1 Strategy 3: B D1 (0.45) \$10 \$0 (0.55) \$10 (0.45) B U3 (0.55) \$0

2. Risk Profiles and Cumulative Risk Profiles (Cont.)
Conclusion: No stochastic dominance exists Strategy A-A1 Strategy A-A2 Strategy B Cumulative Risk Profiles