347 Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permis
Views 157 Downloads 1 File size 336KB
347
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees salesmen. On the other hand, if the new line is handled by the existing sales force, using the existing facilities, the initial investment would only be Rs 30,000, principally for training his present salesmen. The new product sells for Rs 250. The representative normally receives 20 per cent of the sales price on each unit sold, of which 10 per cent is paid as commission to handle the new product. The manufacturer offers to pay 60 per cent of the sale price of each unit sold to the representative, if the representative sets up a separate sales organization. Otherwise the normal 20 per cent will be paid. In either case the salesman gets a 10 per cent commission. Based on the size of the
territory and their experience with other products, the representative estimates the following probabilities for annual sales of the new product: Sales (in units) : Probability :
1,000 0.10
2,000 0.15
3,000 0.40
4,000 0.30
5,000 0.05
(a) Set up a regret table. (b) Find the expected regret of each course of action. (c) Which course of action would have been best under the maximin criterion?
HINTS AND ANSWERS 1. 2. 3. 4.
(i) S2, (ii) S2 or S3, (iii) S3, (iv) S1, (v) S2 (a) Full, (b) Minimal, (c) Full or partial, (d) Partial (a) S3; Rs 150 (b) S3; Rs 80 (c) S1; Rs 40 (d) S3; Rs 55 Choose A: Rs 120, Choose B: Rs 300, Choose C: Rs 176.6, Choose C: Rs 50 5. Let S1 = install new sales facilities Let S2 = continue with existing sales facilities.
11.5
Therefore, payoff function corresponding to S1 and S2 would be S1 = –1,00,000 + 250 × {(30 – 10)/100} × α = –1,00,000 + 50α S2 = –30,000 + 250 × {(20 – 10)/100} × α = – 30,000 + 25α Equating the two, we get – 1,00,000 + 50α = – 30,000 + 25α or α = 2,800.
DECISION-MAKING UNDER RISK
In this decision-making environment, decision-maker has sufficient information to assign probability to the likely occurrence of each outcome (state of nature). Knowing the probability distribution of outcomes (states of nature), the decision-maker needs to select a course of action resulting a largest expected (average) payoff value. The expected payoff is the sum of all possible weighted payoffs resulting from choosing a decision alternative. The widely used criterion for evaluating decision alternatives (courses of action) under risk is the Expected Monetary Value (EMV) or Expected Utility.
11.5.1
Expected Monetary Value (EMV)
The expected monetary value (EMV) for a given course of action is obtained by adding payoff values multiplied by the probabilities associated with each state of nature. Mathematically, EMV is stated as follows: m
EMV (Course of action, Sj) = Σ pij pi i =1
where
m = number of possible states of nature p i = probability of occurrence of state of nature, Ni pij = payoff associated with state of nature Ni and course of action, Sj
Expected monetary value is obtained by adding payoffs for each course of action, multiplied by the probabilities associated with each state of nature.
The Procedure 1. 2. 3.
Construct a payoff matrix listing all possible courses of action and states of nature. Enter the conditional payoff values associated with each possible combination of course of action and state of nature along with the probabilities of the occurrence of each state of nature. Calculate the EMV for each course of action by multiplying the conditional payoffs by the associated probabilities and adding these weighted values for each course of action. Select the course of action that yields the optimal EMV.
Example 11.5 Mr X flies quite often from town A to town B. He can use the airport bus which costs Rs 25 but if he takes it, there is a 0.08 chance that he will miss the flight. The stay in a hotel costs Rs 270 with a 0.96 chance of being on time for the flight. For Rs 350 he can use a taxi which will make 99 per cent chance of being on time for the flight. If Mr X catches the plane on time, he will conclude a business transaction that will produce a profit of Rs 10,000, otherwise he will lose it. Which mode of transport should Mr X use? Answer on the basis of the EMV criterion.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
348 Operations Research: Theory and Applications Solution
Computation of EMV associated with various courses of action is shown in Table 11.3. Courses of Action
States of Nature
Catches the flight
Bus Cost
10,000 – 25 0.92 = 9,975
Miss the flight
Table 11.3
– 25
Stay in Hotel
Prob. Expected Value
0.08
Expected monetary value (EMV)
Cost
Prob.
9,177
10,000 – 270 = 9,730
0.96
– 2.0
– 270
0.04
Taxi
Expected Value
Cost
Prob.
Expected Value
9,340.80
10,000 – 350 = 9,650
0.99
9,553.50
– 10.80
– 350
0.01
– 3.50
9,175
9,330
9,550
Since EMV associated with course of action ‘Taxi’ is largest (= Rs 9,550), it is the logical alternative. Example 11.6 The manager of a flower shop promises its customers delivery within four hours on all flower orders. All flowers are purchased on the previous day and delivered to Parker by 8.00 am the next morning. The daily demand for roses is as follows. Dozens of roses : Probability :
70 0.1
80 0.2
90 0.4
100 0.3
The manager purchases roses for Rs 10 per dozen and sells them for Rs 30. All unsold roses are donated to a local hospital. How many dozens of roses should Parker order each evening to maximize its profits? What is the optimum expected profit? [Delhi Univ., MBA, Dec. 2004] Solution The quantity of roses to be purchased per day is considered as ‘course of action’ and the daily demand of the roses is considered as a ‘state of nature’ because demand is uncertain with known probability. From the data, it is clear that the flower shop must not purchase less than 7 or more than 10 dozen roses, per day. Also each dozen roses sold within a day yields a profit of Rs (30 – 10) = Rs 20 and otherwise it is a loss of Rs 10. Thus Marginal profit (MP) = Selling price – Cost = 30 – 10 = Rs 20 Marginal loss (ML) = Loss on unsold roses = Rs 10 Using the information given in the problem, the various conditional profit (payoff) values for each combination of decision alternatives and state of nature are given by Conditional profit = MP × Roses sold – ML × Roses not sold
=
RS 20D, T 20D − 10(S − D) = 30D − 10S ,
if D ≥ S if D < S
where D = number of roses sold within a day and S = number of roses stocked. The resulting conditional profit values and corresponding expected payoffs are computed in Table 11.4. States of Nature (Demand per Day)
Table 11.4 Conditional Profit Value (Payoffs)
Probability
Conditional Profit (Rs) due to Courses of Action (Purchase per Day)
Expected Payoff (Rs) due to Courses of Action (Purchase per Day)
(1)
70 (2)
80 (3)
90 (4)
100 (5)
70 (1)×(2)
80 (1)×(3)
90 (1)×(4)
100 (1)×(5)
70
0.1
140
130
120
110
14
13
12
11
80
0.2
140
160
150
140
28
32
30
28
90
0.4
140
160
180
170
56
64
72
68
100
0.3
140
160
180
200
42
48
54
60
140
157
168
167
Expected monetary value (EMV)
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
349
Since the highest EMV of Rs 168 corresponds to the course of action 90, the flower shop should purchase nine dozen roses everyday. Example 11.7 A retailer purchases cherries every morning at Rs 50 a case and sells them for Rs 80 a case. Any case that remains unsold at the end of the day can be disposed of the next day at a salvage value of Rs 20 per case (thereafter they have no value). Past sales have ranged from 15 to 18 cases per day. The following is the record of sales for the past 120 days. Cases sold : Number of days :
15 12
16 24
17 48
18 36
Find out how many cases should the retailer purchase per day in order to maximize his profit. [Delhi Univ., MCom, 2000; Ajmer Univ., MBA, 2003]
Solution Let Ni (i = 1, 2, 3, 4) be the possible states of nature (daily likely demand) and Sj ( j = 1, 2, 3, 4) be all possible courses of action (number of cases of cherries to be purchased). Marginal profit (MP) = Selling price – Cost = Rs (80 – 50) = Rs 30 Marginal loss (ML) = Loss on unsold cases = Rs (50 – 20) = Rs 30 The conditional profit (payoff) values for each combination of decision alternatives and state of nature are given by Conditional profit = MP × Cases sold – ML × Cases unsold = (80 – 50) (Cases sold) – (50 –20) (Cases unsold) 30S = (80 – 50) S – 30(N – S) = 60S – 30N
if S ≥ N if S < N
The resulting conditional profit values and corresponding expected payoffs are computed in Table 11.5. States of Nature (Demand per Week)
Probability
(1)
Conditional Profit (Rs) due to Courses of Action (Purchase per Day) 15 (2)
16 (3)
17 (4)
Expected Payoff (Rs) due to Courses of Action (Purchase per Day)
18 (5)
15 (1)×(2)
16 (1)×(3)
17 (1)×(4)
18 (1)×(5)
15
0.1
450
420
390
360
45
42
39
36
16
0.2
450
480
450
420
90
96
90
84
17
0.4
450
480
510
480
180
192
204
192
18
0.3
450
480
510
540
135
144
153
162
450
474
486
474
Expected monetary value (EMV)
Table 11.5 Conditional Profit Value (Payoffs)
Since the highest EMV of Rs 486 corresponds to the course of action 17, the retailer must purchase 17 cases of cherries every morning. Example 11.8 The probability of demand for hiring cars on any day in a given city is as follows: No. of cars demanded Probability
: :
0 0.1
1 0.2
2 0.3
3 0.2
4 0.2
Cars have a fixed cost of Rs 90 each day to keep the daily hire charges (variable costs of running) Rs 200. If the car-hire company owns 4 cars, what is its daily expectation? If the company is about to go into business and currently has no car, how many cars should it buy? Solution Given that Rs 90 is the fixed cost and Rs 200 is variable cost. The payoff values with 4 cars at the disposal of decision-maker are calculated as under: No. of cars demanded : 0 1 2 3 4 Payoff : 0 – 90 × 4 200 – 90 × 4 400 – 90 × 4 600 – 90 × 4 800 – 90 × 4 (with 4 cars) = – 360 = – 160 = 40 = 240 = 440
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
350 Operations Research: Theory and Applications Thus, the daily expectation is obtained by multiplying the payoff values with the given corresponding probabilities of demand: Daily Expectation = (– 360)(0.1) + (– 160)(0.2) + (40)(0.3) + (240)(0.2) + (440)(0.2) = Rs 80 The conditional payoffs and expected payoffs for each course of action are shown in Tables 11.6 and 11.7. Demand of Cars
0 1 2 3 4
Table 11.6 Conditional Payoff Values
0.1 0.2 0.3 0.2 0.2
Demand of Cars
0 1 2 3 4
Table 11.7 Expected Payoffs and EMV
Probability
0.1 0.2 0.3 0.2 0.2
EMV
Conditional Payoff (Rs) due to Decision to Purchase Cars (Course of Action)
Probability 0
1
2
3
0 0 0 0 0
– 90 110 110 110 110
– 180 20 220 220 220
– 270 – 70 130 330 330
4 – 360 – 160 40 240 440
Conditional Payoff (Rs) due to Decision to Purchase Cars (Course of Action) 0
1
2
3
4
0 0 0 0 0
–9 22 33 22 22
– 18 4 66 44 44
– 27 – 14 39 66 66
– 36 – 32 12 48 88
0
90
140
130
80
Since the EMV of Rs 140 for the course of action 2 is the highest, the company should buy 2 cars.
11.5.2
Expected Opportunity Loss (EOL)
Expected opportunity loss (EOL), also called expected value of regret, is an alternative decision criterion for decision making under risk. The EOL is defined as the difference between the highest profit (or payoff ) and the actual profit due to choosing a particular course of action in a particular state of nature. Hence, EOL is the amount of payoff that is lost by not choosing a course of action resulting to the minimum payoff in a particular state of nature. A course of action resulting to the minimum EOL is preferred. Mathematically, EOL is stated as follows. m
EOL (State of nature, Ni ) = Σ lij pi i =1
where
lij = opportunity loss due to state of nature, Ni and course of action, Sj p i = probability of occurrence of state of nature, Ni
The Procedure 1. Prepare a conditional payoff values matrix for each combination of course of action and state of nature along with the associated probabilities. 2. For each state of nature calculate the conditional opportunity loss (COL) values by subtracting each payoff from the maximum payoff. 3. Calculate the EOL for each course of action by multiplying the probability of each state of nature with the COL value and then adding the values. 4. Select a course of action for which the EOL is minimum. Example 11.9 A company manufactures goods for a market in which the technology of the product is changing rapidly. The research and development department has produced a new product that appears to have potential for commercial exploitation. A further Rs 60,000 is required for development testing. The company has 100 customers and each customer might purchase, at the most, one unit of the product. Market research suggests that a selling price of Rs 6,000 for each unit, with total variable costs of manufacturing and selling estimate as Rs 2,000 for each unit.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
351
From previous experience, it has been possible to derive a probability distribution relating to the proportion of customers who will buy the product as follows: Proportion of customers : 0.04 0.08 0.12 0.16 0.20 Probability : 0.10 0.10 0.20 0.40 0.20 Determine the expected opportunity losses, given no other information than that stated above, and state whether or not the company should develop the product. Solution If p is the proportion of customers who purchase the new product, the company’s conditional profit is: (6,000 – 2,000) × 100 p – 60,000 = Rs (4,00,000 p – 60,000). Let Ni (i = 1, 2, . . ., 5) be the possible states of nature, i.e. proportion of the customers who will buy the new product and S1 (develop the product) and S2 (do not develop the product) be the two courses of action. The conditional profit values (payoffs) for each pair of Nis and Sjs are shown in Table 11.8. Proportion of Customers (State of Nature)
0.04 0.08 0.12 0.16 0.20
Conditional Profit = Rs (4,00,000 p – 60,000) Course of Action S1 (Develop)
S2 (Do not Develop)
– 44,000 – 28,000 – 12,000 4,000 20,000
0 0 0 0 0
Table 11.8 Conditional Profit Values (Payoffs)
Opportunity loss values are shown in Table 11.9. Proportion of Customers (State of Nature)
Probability
0.04 0.08 0.12 0.16 0.20
0.1 0.1 0.2 0.4 0.2
Conditional Profit (Rs)
Opportunity Loss (Rs)
S1
S2
S1
S2
– 44,000 – 28,000 – 12,000 4,000 20,000
0 0 0 0 0
44,000 28,000 12,000 0 0
0 0 0 4,000 20,000
Using the given estimates of probabilities associated with each state of nature, the expected opportunity loss (EOL) for each course of action is given below: EOL (S1) = 0.1 (44,000) + 0.1 (28,000) + 0.2 (12,000) + 0.4 (0) + 0.2 (0) = Rs 9,600 EOL (S2) = 0.1 (0) + 0.1 (0) + 0.2 (0) + 0.4 (4,000) + 0.2 (20,000) = Rs 5,600 Since the company seeks to minimize the expected opportunity loss, the company should select course of action S2 (do not develop the product) with minimum EOL.
11.5.3
Expected Value of Perfect Information (EVPI)
If decision-makers can get perfect (complete and accurate) information about the occurrence of various states of nature, then choosing a course of action that yields the desired payoff in the presence of any state of nature is easy. The EMV or EOL criterion helps the decision-maker to select a particular course of action that optimizes the expected payoff, without any additional information. Expected value of perfect information (EVPI) represents the maximum amount of money required to pay for getting additional information about the occurrence of various states of nature before arriving to a decision. Mathematically, it is stated as: EVPI = (Expected profit with perfect information) – (Expected profit without perfect information) =
Table 11.9 Opportunity Loss Values
Expected value of perfect information is an average (or expected) value of an additional information if it were of any worth.
m
pij ) – EMV* ∑ pi max( j
i =1
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
352 Operations Research: Theory and Applications where
EVPI provides an instant way to check whether getting any additional information might be worthwhile.
pij = best payoff when action, Sj is taken in the presence of state of nature, Ni p i = probability of state of nature, Ni EMV* = maximum expected monetary value
Example 11.10 A company needs to increase its production beyond its existing capacity. It has narrowed down on two alternatives in order to increase the production capacity: (a) expansion, at a cost of Rs 8 million, or (b) modernization at a cost of Rs 5 million. Both approaches would require the same amount of time for implementation. Management believes that over the required payback period, demand will either be high or moderate. Since high demand is considered to be somewhat less likely than moderate demand, the probability of high demand has been set at 0.35. If the demand is high, expansion would gross an estimated additional Rs 12 million but modernization would only gross an additional Rs 6 million, due to lower maximum production capability. On the other hand, if the demand is moderate, the comparable figures would be Rs 7 million for expansion and Rs 5 million for modernization. (a) Calculate the conditional profit in relation to various action-and-outcome combinations and states of nature. (b) If the company wishes to maximize its expected monetary value (EMV), should it modernize or expand? (c) Calculate the EVPI. (d) Construct the conditional opportunity loss table and also calculate EOL. [Delhi Univ, MBA, 2004] Solution (a) States of nature: High demand and Moderate demand, and Courses of action: Expand and Modernize. Since probability of high demand is estimated at 0.35, the probability of moderate demand must be (1 – 0.35) = 0.65. The calculations for conditional profit values are shown in Table 11.10. State of Nature (Demand)
Conditional Profit (million Rs) due to Course of Action Expand (S1)
Table 11.10 Conditional Profit Table
High demand (N1)
12 – 8 =
Moderate demand (N2)
Modernize (S2)
4
6 – 5 = 1
7 – 8 = –1
5 – 5 = 0
(b) The payoff table (Table 11.10) can be rewritten as follows along with the given probabilities of states of nature. State of Nature (Demand)
Probability
Conditional Profit (million Rs) Due to Course of Action Expand
Table 11.11 Payoff Table
Modernize
High demand
0.35
4
1
Moderate demand
0.65
–1
0
The calculation of EMV for each course of action S1 and S2 is given below: EMV(S1) = 0.35(4) + 0.65( – 1) = Rs 0.75 million EMV(S2) = 0.35(1) + 0.65(0) = Rs 0.35 million Since EMV (S1) = 0.75 million is maximum, the company must choose course of action S1(expand). (c) To calculate EVPI, first calculate EPPI by choosing optimal course of action for each state of nature. Multiply conditional profit associated with each course of action by the given probability to get weighted profit, and then add these weights as shown in Table 11.12. State of Nature (Demand)
High demand Moderate demand Table 11.12
Probability
0.35 0.65
Optimal Course of Action
S1 S2
Profit from Optimal Course of Action (Rs million) Conditional Profit
Weighted Profit
4 0
4 × 0.35 = 1.40 0 × 0.65 = 0 EPPI = 1.40
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
If optimal EMV* = Rs 0.75 million corresponding to the course of action S1, then EVPI = EPPI – EMV(S1) = 1.40 – 0.75 = Rs 0.65 million. In other words, to get perfect information on demand pattern (high or moderate), company should consider paying up to Rs 0.65 million. (d) The opportunity loss values are shown in Table 11.13. State of Nature (Demand)
Probability
High demand (N1) Moderate demand (N2)
0.35 0.65
Conditional Profit (Rs million) Due to Course of Action
Conditional Opportunity Loss (Rs million) Due to Course of Action
S1
S2
S1
S2
4 –1
1 0
0 1
3 0
Table 11.13
Since probabilities associated with each state of nature, P (N1) = 0.35, and P(N2) = 0.65, the expected opportunity losses for the two courses of action are: EOL(S1) = 0.35(0) + 0.65(1) = Re 0.65 million EOL(S2) = 0.35(3) + 0.65(0) = Rs 1.05 million Since the expected opportunity loss, EOL (S1) = Re 0.65 million minimum, decision-maker must select course of action S1, so as to have smallest expected opportunity loss. Example 11.11 A certain piece of equipment has to be purchased for a construction project at a remote location. This equipment contains an expensive part that is subject to random failure. Spares of this part can be purchased at the same time the equipment is purchased. Their unit cost is Rs 1,500 and they have no scrap value. If the part fails on the job and no spare is available, the part will have to be manufactured on a special order basis. If this is required, the total cost including down time of the equipment, is estimated at Rs 9,000 for each such occurrence. Based on previous experience with similar parts, the following probability estimates of the number of failures expected over the duration of the project are provided below: Failure : 0 1 2 Probability : 0.80 0.15 0.05 (a) Determine optimal EMV* and optimal number of spares to purchase initially. (b) Based on opportunity losses, determine the optimal course of action and optimal value of EOL. (c) Determine the expected profit with perfect information and expected value of perfect information. Solution (a) Let N1 (no failure), N2 (one failure) and N3 (two failures) be the possible states of nature (i.e. number of parts failures or number of spares required). Similarly, let S1 (no spare purchased), S2 (one spare purchased) and S3 (two spares purchased) be the possible courses of action. The conditional costs for each pair of course of action and state of nature is shown in Table 11.14. State of Nature (Spare Required)
Course of Action (Number of Spare Purchased)
Purchase Cost (Rs)
Emergency Cost (Rs)
Total Conditional Cost (Rs)
0 N1 0 0
0
0
0
0
1
1,500
0
1,500
2
3,000
0
3,000
1 N2 1 1
0
0
9,000
9,000
1
1,500
0
1,500
2
3,000
0
3,000
2 N3 2 2
0
0
18,000
18,000
1
1,500
9,000
10,500
2
3,000
0
3,000
Table 11.14
Using the conditional costs and the probabilities of states of nature, the expected monetary value can be calculated for each of three states of nature as shown in Table 11.15.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
353
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
354 Operations Research: Theory and Applications State of Nature (Space Required)
Probability
Conditional Cost Due to Course of Action S1
N1
0.80
0
Weighted Cost Due to Course of Action
S2
S3
S1
S2
S3
1,500
3,000
0.80 (0)
1,200
2,400
225
450
= 900
525
150
EMV = 2,250
1,950
3,000
= 0
Table 11.15 Expected Monetary Value
N2
0.15
9,000
1,500
3,000
0.15 (9,000) = 1,350
N3
0.05
18,000
10,500
3,000
0.05 (18,000)
Since weighted cost = Rs 1,950 is lowest due to course of action, S2, it should be chosen. If the EMV is expressed in terms of profit, then EMV* = EMV(S2) = – Rs 1,950. Hence, the optimal number of spares to be purchased initially should be one. (b) The calculations for conditional opportunity loss (COL) to determine EOL are shown in Table 11.16. State of Nature (Space Required) Table 11.16 Conditional Opportunity Loss (COL)
Conditional Cost Due to Course of Action
N1 N2 N3
Conditional Opportunity Loss Due to Course of Action
S1
S2
S3
S1
0 9,000 18,000
1,500 1,500 10,500
3,000 3,000 3,000
0 7,500 15,000
S2
S3
1,500 0 7,500
3,000 1,500 0
Since we are dealing with conditional costs rather than conditional profits, the lower value for each state of nature shall be considered for calculating opportunity losses. The calculations for expected opportunity loss are shown in Table 11.17. State of Nature (Space Required)
Probability
Conditional Opportunity Loss (Cost) Due to Course of Action S1
Table 11.17 Expected Opportunity Loss (EOL)
Weighted Opportunity Loss (Cost) Due to Course of Action
S2
S3
S1
N1
0.80
0
1,500
3,000
N2
0.15
7,500
0
1,500
N3
0.05
15,000
7,500
0
S2
S3
0.80(0) = 0 0.15 (7,500) = 1,125 0.05 (15,000) = 750
1,200
2,400
0
225
375
0
EMV = 1,875
1,575
2,625
Since minimum, EOL* = EOL(S2) = Rs 1,575, therefore adopt course of action S2 and purchase one spare. (c) The expected profit with perfect information (EPPI) can be determined by selecting the optimal course of action for each state of nature, multiplying its conditional values by the corresponding probability and then adding these products. The EPPI calculations are shown in Table 11.18. States of Nature
Probability
Optimal Course of Action
(Space Required) N1 N2 N2 Table 11.18
0.80 0.15 0.05
S1 S2 S3
Cost of Optimal Course of Action (Rs) Conditional Cost (Minimum Value)
Weighted Opportunity Loss
0 1,500 3,000
0.80(0) = 0 0.15 (1,500) = 225 0.05 (3,000) = 150 Total = 375
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
355
Thus expected profit with perfect information is, EPPI = Rs 375. Expected value of perfect information then is: EVPI = EPPI – EMV* = – 375 – (– 1,950) = Rs 1,575. It may be observed that, EVPI = EOL* = Rs 1,575 Example 11.12 XYZ Company manufactures parts for passenger cars and sells them in lots of 10,000 parts each. The company has a policy of inspecting each lot before it is actually shipped to the retailer. Five inspection categories, established for quality control, represent the percentage of defective items contained in each lot. These are given in the following table. The daily inspection chart for past 100 inspections shows the following rating or breakdown inspection: Due to this the management is considering two possible courses of action: (i) S1: Shut down the entire plant operations and thoroughly inspect each machine. Rating Excellent (A) Good (B) Acceptable (C) Fair (D) Poor (E)
Proportion of Defective Items
Frequency
0.02 0.05 0.10 0.15 0.20
25 30 20 20 5 Total = 100
(ii) S2 : Continue production as it now exists but offer the customer a refund for defective items that S2 : are discovered and subsequently returned. The first alternative will cost Rs 600 while the second alternative will cost the company Re 1 for each defective item that is returned. What is the optimum decision for the company? Find the EVPI. Solution Rating
A B C D E
Calculations of inspection and refund cost are shown in Table 11.19. Defective Rate
Probability
0.02 0.05 0.10 0.15 0.20
Cost
Opportunity Loss
Inspect
Refund
Inspect
Refund
0.25 0.30 0.20 0.20 0.05
600 600 600 600 600
200 500 1,000 1,500 2,000
400 100 0 0 0
0 0 400 900 1,400
1.00
600*
670
EOL = 170*
240
Table 11.19 Inspection and Refund Cost
The cost of refund is calculated as follows: For lot A : 10,000 × 0.02 × 1.00 = Rs 200 Similarly, the cost of refund for other lots is calculated. Expected cost of refund is: 200 × 0.25 + 500 × 0.30 + . . . + 2,000 × 0.05 = Rs 670 Expected cost of inspection is: 600 × 0.25 + 600 × 0.30 + . . . + 600 × 0.05 = Rs 600 Since the cost of refund is more than the cost of inspection, the plant should be shut down for inspection. Also, EVPI = EOL of inspection = Rs 170. Example 11.13 A toy manufacturer is considering a project of manufacturing a dancing doll with three different movement designs. The doll will be sold at an average of Rs 10. The first movement design using ‘gears and levels’ will provide the lowest tooling and set up cost of Rs 1,00,000 and Rs 5 per unit of variable cost. A second design with spring action will have a fixed cost of Rs 1,60,000 and variable cost of Rs 4 per unit. Yet another design with weights and pulleys will have a fixed cost of Rs 3,00,000 and variable cost Rs 3 per unit. The demand events that can occur for the doll and the probability of their occurrence is given below:
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
356 Operations Research: Theory and Applications Demand (units)
Probability
25,000 1,00,000 1,50,000
0.10 0.70 0.20
Light demand Moderate demand Heavy demand
(a) Construct a payoff table for the above project. (b) Which is the optimum design? (c) How much can the decision-maker afford to pay in order to obtain perfect information about the demand? Solution
The calculations for EMV are shown in Table 11.20. Payoff = (Demand × Selling price) – (Fixed cost + Demand × Variable cost) = Revenue – Total variable cost – Fixed cost
States of Nature (Demand)
Table 11.20 EMV and Payoff Values
Light Moderate Heavy
Probability
0.10 0.70 0.20
Conditional Payoff (Rs) Due to Courses of Action (Choice of Movements)
Expected Payoff (Rs) Due to Courses of Action
Gears and Levels
Spring Action
Weights and Pulleys
Gears and Levels
Spring Action
Weights and Pulleys
25,000 4,00,000 6,50,000
– 10,000 4,40,000 7,40,000
– 1,25,000 4,00,000 7,50,000
2,500 2,80,000 1,30,000
– 1,000 3,08,000 1,48,000
– 12,500 2,80,000 1,50,000
4,12,500
4,55,000
4,17,500
EMV
Since EMV is largest for spring action, it is the one that must be selected.
Table 11.21 Expected Payoff with Perfect Information
States of Nature (Demand)
Probability
Light Moderate Heavy
0.10 0.70 0.20
Courses of Action Gears and Spring Weights Levels Action and Pulleys 25,000 4,00,000 6,50,000
– 10,000 4,40,000 7,40,000
– 1,25,000 4,00,000 7,50,000
Maximum Payoff
Maximum Payoff × Probability
25,000 4,40,000 7,50,000
2,500 3,08,000 1,50,000 Total = 4,60,500
The maximum amount of money that the decision-maker would be willing to pay in order to obtain perfect information regarding demand for the doll will be EVPI = Expected payoff with perfect information – Expected payoff under uncertainty (EMV) = 4,60,500 – 4,55,000 = Rs 5,500 Example 11.14 A TV dealer finds that the cost of holding a TV in stock for a week is Rs. 50. Customers who cannot obtain new TV sets immediately tend to go to other dealers and he estimates that for every customer who cannot get immediate delivery he loses an average of Rs. 200. For one particular model of TV the probabilities of demand of 0, 1, 2, 3, 4 and 5 TV sets in a week are 0.05, 0.10, 0.20, 0.30, 0.20 and 0.15, respectively. (a) How many televisions per week should the dealer order? Assume that there is no time lag between ordering and delivery. (b) Compute EVPI. (c) The dealer is thinking of spending on a small market survey to obtain additional information regarding the demand levels. How much should he be willing to spend on such a survey. [Delhi Univ., MBA, 2000] Solution If D denotes the demand and S the number of televisions stored (ordered), then the conditional cost values are computed and are shown in Table 15.22. 50S + 200 ( D − S ) , when D ≥ S
Cost function = 50 D + 50 ( S − D ) , when D < S
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees State of Nature (Demand)
Probability
Conditional Cost (Rs.) Course of Action (Stock) 0
1
2
3
4
5
0
0.05
0
50
100
150
200
250
1
0.10
200
50
100
150
200
250
2
0.20
400
250
100
150
200
250
3
0.30
600
450
300
150
200
250
4
0.20
800
650
500
350
200
250
5
0.15
1000
850
700
550
400
250
Expected Cost = 590
450
330
250
230
250
State of Nature (Demand)
357
Probability (1)
Minimum Cost for Perfect Information (2)
Table 11.22 Expected Cost
Expected Cost for Perfect Information (3) = (1) × (2)
0
0.05
0
1
0.10
50
0 5
2
0.20
100
20
3
0.30
150
45
4
0.20
200
40
5
0.15
250
37.5
Table 11.23 Expected Payoff with perfect Information
ECPI = 147.5
EVPI = Conditional Cost – ECPI = 230 – 147.5 = Rs. 82.5. The pay-off for EMV is shown in Table 15.24 State of Nature (Demand)
0
Probability
Conditioinal Payoff Course of Action (Stock)
0.05
0
1
2
3
4
5
0
– 50
– 100
– 150
– 200
– 250
1
0.10
0
150
100
50
0
– 50
2
0.20
0
150
300
250
200
150
3
0.30
0
150
300
450
400
350
4
0.20
0
150
300
450
600
550
5
0.15
0
150
300
450
600
750
EMV = 0
140
260
340
360
340
State of Nature (Demand) 0 1 2 3 4 5
Probability (1)
Payoff for Perfect Information (2)
Expected Payoff for Perfect Information (3) = (1) × (2)
0.05 0.10 0.20 0.30 0.20 0.15
0 150 300 450 600 750
0 15 60 135 120 112.5 442.5
Table 11.24 Computation of EMV
Table 11.25 Computation of EPPI
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
358 Operations Research: Theory and Applications The expected value of perfect information is given by EVPI = EPPI – EMV* = 442.5 – 360 = Rs. 82.5. (c)
On the basis of the given data, the dealer should not be willing to spend more than Rs. 82.5 for the market survey.
Example 11.15 XYZ company is considering issuing 1,00,000 shares to raise capital needed for expansion. It is estimated that if the issues were made now, it would be fully taken up at a price of Rs. 30 per share. However, the company is facing two crucial situations, both of which may influence the share price in the near future, namely: (i) A wage dispute with tool room operators which could lead to a strike and have an adverse effect on the share price. (ii) The possibility of a substantial contract with a large company overseas which would increase the share price. The four possible outcomes and their expected effect on the company’s share prices are: E1 : No strike and contract obtained–share price rises to Rs. 34. E2 : Strike and contract obtained–share price stays at Rs. 30. E3 : No strike and contract lost–share price raises to Rs. 32. E4 : Strike and contract lost–share price falls to Rs. 16. Management has identified three possible strategies that the company could adopt, namely A1 : Issue 1,00,000 shares now A2 : Issue 1,00,000 shares only after the outcomes of (i) and (ii) are known A3 : Issue 5,00,000 shares now and 50,000 shares after the outcomes of (i) and (ii) are known. (a) Determine the maximax solution. What alternative criterion might be used? (b) It has been estimated that the probability of a strike is 55 per cent and that there is a 65 per cent chance of getting the contract, these probabilities being independent. Determine the optimum policy for the company using the criterion of maximizing expected pay-off. (c) Determine the expected value of perfect information for the company. Solution States of Nature E1 E2 E3 E4 Table 11.26
The payoff values are shown in Table 11.26 Probability
(1 – 0.55) × 0.65 0.55 × 0.65 (1 – 0.55)(1 – 0.65) 0.55 × (1 – 0.65)
Conditional Payoff (Rs. lakh) Alternative Strategy A1 A2 A3 = 0.2925 = 0.3575 = 0.1575 = 0.1925
Expected Values
30 30 30 30
34 30 32 16
30
34
0.5 × 30 + 0.5 × 34 0.5 × 30 + 0.5 × 30 0.5 × 30 + 0.5 × 32 0.5 × 30 + 0.5 × 16
Conditional Loss (Rs. lakh) Alternative Strategy A1 A2 A3 = 31 = 30 = 30.5 = 23
4 0 2 0
0 0 0 14
3 0 1.5 7
31
A strategy with highest minimum (maximin) payoff (i.e. 30) is A1 and a strategy with highest (maximax) payoff (i.e. 34) is A2. Since highest pay-off of Rs. 30 lakh is obtained corresponding to strategy A1, the company should adopt strategy A1. (c) Calculations for expected value of the perfect information are shown in Table 15.27.
Table 11.27 EVPI
State of Nature
Maximum Pay-off
Probability
Expected Pay-off with Perfect Information
E1 E2 E3 E4
34 30 32 16
0.2925 0.3575 0.1575 0.1925
9.94 10.72 5.04 5.78 31.48
Hence, expected value of perfect information is : 31.48 – 30 = Rs. 1,48,000.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
Example 11.16 A vegetable seller buys tomatoes for Rs. 45 a box and sells them for Rs. 80 per box. If the box is not sold on the first selling day, it is worth Rs. 15 as salvage. The past records indicate that demand is normally distriubuted, with a mean of 30 boxes daily and a standard deviation of 9 boxes. How many boxes should he stock? Solution The probability of selling at least one additional unit (box) to justify the stocking of additional unit is given by IL (45 − 15) p= = = 30 = 0.462 IP + IL (80 − 45) + (45 − 15) 65 where Incremental loss (IL) = Cost price – Salvage price Incremental profit (IP) = Selling price – Cost price This implies that the vegetable seller must be 46.2 per cent sure of selling at least one additional unit before he should pay to stock an additional unit. The vegetable seller should stock additional boxes until point A is reached. If more units are stocked, then the probability will fall below 0.462. The point A is at 0.1 standard deviation to the right of the mean x = 30. Since the standard deviation of the distribution of past demand is 9 boxes, point A can be located as follows: Point A = Mean + Standard deviation = 30 + 0.1 × 9 = 30.1 31 boxes Hence, the fruit seller should stock 31 boxes.
Fig. 11.1
Example 11.17 A stall at a certain railway station sells for Rs. 1.50 paise a copy of daily newspaper for which it pays Rs. 1.20. Unsold papers are returned for a refund of Re. 0.95 a copy. Daily sales and corresponding probabilities are as follows : Daily sales : 500 600 700 Probability : 0.5 0.3 0.2 (a) How many copies should it order each day to get maximum expected profit? (b) If unsold copies cannot be returned and are useless, what should be the optimal order each day? Use increment analysis. Solution Given that, Incremental profit (IP) = Rs. (1.50 – 1.20) = Re. 0.30 Incremental loss (IL) = Rs. (1.20 – 0.95) = Re. 0.25 The probability (p) of selling at least one additional copy of the newspaper to justify keeping that additional copy of newspaper is: IL 0.25 = = 0.45. p= IP + IL 0.30 + 0.25 Thus to justify the ordering of an additional copy, there must be at least 0.45 cumulative probability of selling that copy. Cumulative probabilities are computed below: Daily sales : 500 600 700 Probability : 0.50 0.30 0.20 Comulative probability : 1.00 0.50 0.20 Hence, the optimal order size is 600 copies. (b) If unsold copies are non-refundable, then IL 1.20 1.20 = = = 0.80 p= IP + IL 0.30 + 1.20 1.50 Hence, optimal order size is, 500 copies. Example 11.18 The demand pattern of the cakes made in a bakery is as follows: No. of cakes demanded : 0 1 2 3 4 5 Probability : 0.05 0.10 0.25 0.30 0.20 0.10 If the preparation cost is Rs 30 per unit and selling price is Rs 40 per unit, how many cakes should the baker bake for maximizing his profit?
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
359
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
360 Operations Research: Theory and Applications Solution Given that incremental cost (IC) to prepare a cake is Rs 30 per unit and incremental price (IP) to sell a cake is Rs 40 per unit. The cumulative probability ( p) of selling at least an additional unit of cake to justify the stocking of that additional unit of cake is given by p=
30 IC = = 0.428 IC + IP 30 + 40
The cumulative probabilities of greater than type are computed as shown in Table 11.28. Demand ( No. of Cakes)
Probability P (Demand = k)
0 1 2 3 4 5
0.05 0.10 0.25 0.30 0.20 0.10
Table 11.28
Cumulative Probability P (Demand ≥ k) 1.00 0.95 0.85 0.60 ← 0.30 0.10
Since P(demand ≥ k) that exceeds the critical ratio, p = 0.428 is k = 3 units of cake, the optimal decision is to prepare only 3 cakes.
11.6
Posterior probabilities are the revised probabilities of the states of nature obtained after conducting a test to improve the prior probabilities of respective nature.
POSTERIOR PROBABILITIES AND BAYESIAN ANALYSIS
An initial probability statement to evaluate expected payoff is called a prior probability distribution, but if the probability statement has been revised due to additional information, then such a probability statement is called a posterior probability distribution. In this section we will discuss the method of computing posterior probabilities, given prior probabilities using Bayes’ theorem. The analysis of problems using posterior probabilities with new expected payoffs and additional information, is called prior-posterior analysis.
Baye’s Theorem Statement Let A1, A2, . . ., An be mutually exclusive and collectively exhaustive outcomes. Their probabilities P(A1), P(A2), . . ., P(An) are known. There is an experimental outcome B for which the conditional probabilities P(B | A1), P(B | A2), . . ., P(B | An) are also known. Given the information that outcome B has occurred, the revised conditional probabilities of outcomes Ai, i.e. P(Ai | B), i = 1, 2, . . ., n are determined by using the following relationship: P ( Ai and B ) P ( Ai ∩ B ) P ( Ai ) P ( B | Ai ) P ( Ai | B ) = = = P( B) P( B) P( B) where
P( B) =
n
∑ P( Ai ) P( B | Ai )
i =1
Since each joint probability can be expressed as the product of a known marginal (prior) and conditional probability, i.e., P ( Ai ∩ B ) = P ( Ai ) × P ( B | Ai )
Baye’s decision rule uses the prior probabilities to determine the expected payoff for each decision alternatives and then chooses the one with the largest expected payoff.
Example 11.19 A company is considering to introduce a new product to its existing product range. It has defined two levels of sales as ‘high’ and ‘low’ on which it wants to base its decision and has estimated the changes that each market level will occur, together with their costs and consequent profits or losses. This information is summarized below: States of Nature
Probability
Courses of Action Market the Product (Rs ’000)
Do not Market the Product (Rs ’000)
High sales
0.3
150
0
Low sales
0.7
– 40
0
The company’s marketing manager suggests that a market research survey may be undertaken to provide further information on which the company should base its decision. Based on the company’s past
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Decision Theory and Decision Trees
experience with a certain market research organization, the marketing manager assesses its ability to give good information in the light of subsequent actual sales achievements. This information is given below: Actual Sales
Market Research (Survey outcome)
Market ‘high’
Market ‘low’
‘High’ sales forecast
0.5
0.1
Indecisive survey report
0.3
0.4
‘Low’ sales forecast
0.2
0.1
The market research survey costs Rs 20,000, state whether or not there is a case for employing the market research organization. [Delhi Univ., MBA, 2000] Solution
The expected monetary value (EMV) for each course of action is shown in Table 11.29.
States of Nature High sales (N1) Low sales (N2)
Prior Probability
Courses of Action
Expected Profit (’000 Rs)
Do not Market
Market
Do not Market
150 – 40
0 0
45 – 28
0 0
EMV = 17
= 0
0.3 0.7
Market
Table 11.29
With no additional information, the company should choose course of action ‘market the product’. However, if the company had the perfect information about the ‘low sales’, then company would not go ahead with the decision because expected value would be (–) Rs 28,000. Thus, the value of perfect information is the expected value of low sales. Let outcomes of the research survey be: high sales (S1), indecisive report (S2) and low sales (S3) and states of nature be: high market (N1) and low market (N2). The calculations for prior probabilities of forecast are shown in Table 11.30. Outcome
Sales Prediction High Market ( N1 )
Low Market ( N2 )
High sales (S1) Indecisive report (S2)
P(S1 | N1) = 0.5 P(S2 | N1) = 0.3
P(S1 | N2) = 0.1 P(S2 | N2) = 0.4
Low sales (S3)
P(S3 | N1) = 0.2
P(S3 | N2) = 0.5
Table 11.30
With additional information, the company can now revise the prior probabilities of outcomes to get posterior probabilities. The calculations of the revised probabilities, given the sales forecast are shown in Table 11.31. States of Nature
Prior Probability P(Ni)
Conditional Probability P(Si | Ni)
High sales (N1)
0.3
P(S1 | N1) = 0.5
0.15
–
–
P(S2 | N1) = 0.3
–
0.09
–
P(S3 | N1) = 0.2
–
–
0.06
P(S1 | N2) = 0.1
0.07
–
–
P(S2 | N2) = 0.4
–
0.28
–
P(S3 | N2) = 0.5
–
–
0.35
0.22
0.37
0.41
Low sales (N2)
Marginal Probability
0.7
Joint Probability P(Si ∩ Ni) = P(Ni) P(Si | Ni)
Table 11.31 Revised Probabilities
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102
361
Copyright © 2016. Laxmi Publications Pvt Ltd. All rights reserved. May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
362 Operations Research: Theory and Applications The posterior probabilities of actual sales, given the sales forecast, are: Outcome (Si)
Probability P(Si)
States of Nature (Ni)
Posterior Probability P(Ni | Si) = P(Ni ∩ Si)/P(Si)
S1
0.22
N1 N2
0.15/0.22 = 0.681 0.07/0.22 = 0.318
S2
0.37
N1 N2
0.09/0.37 = 0.243 0.28/0.37 = 0.756
S3
0.41
N1 N2
0.06/0.41 = 0.146 0.35/0.41 = 0.853
Given the additional information, the revised probabilities to calculate net expected value with respect to each outcome are shown in Table 11.32. Sales Forecast States of Nature
Table 11.32
Revised Conditional Profit (Rs)
High
Indecisive
Low
Prob.
EV (Rs)
Prob.
EV (Rs)
Prob.
EV (Rs)
High sales
130
0.681
88.53
0.243
31.59
0.146
18.98
Low sales
– 60
0.318
–19.08
0.756
– 45.36
0.853
– 51.18
Expected value of sales forecast
69.45
– 13.77
– 32.20
Probability of occurrence Net expected value (Expected value × Probability)
0.22
0.37
0.41
15.279
– 5.095
13.202
CONCEPTUAL QUESTIONS B 1. Given the complete set of outcomes in a certain situation, how is the EMV determined for a specific course of action? Explain in your own words. 2. Explain the difference between expected opportunity loss and expected value of perfect information. 3. Indicate the difference between decision-making under risk, and uncertainty, in statistical decision theory.
4. Briefly explain ‘expected value of perfect information’ with examples. 5. Describe a business situation where a decision-maker faces a decision under uncertainty and where a decision based on maximizing the expected monetary value cannot be made. How do you think the decision-maker should make the required decision?
SELF PRACTICE PROBLEMS B 1. You are given the following payoffs of three acts A1, A2 and A3 and the events E1, E2, E3. States of Nature
Three Acts A1
A3
A3
E1
25
– 10
– 125
E2
400
440
400
E3
650
740
750
The probabilities of the states of nature are 0.1, 0.7 and 0.2, respectively. Calculate and tabulate the EMV and conclude which would prove to be the best course of action. 2. The management of a company is faced with the problem of choosing one of three products that it wants to manufacture. The potential demand for each product may turn out to be good, moderate or poor. The probabilities for each of the states of nature were estimated as follows.
Product X Y Z
Nature of Demand Good
Moderate
Poor
0.70 0.50 0.40
0.20 0.30 0.50
0.10 0.20 0.10
The estimated profit or loss in rupees under the three states may be taken as: Product
Good
Moderate
X Y Z
30,000 60,000 40,000
20,000 30,000 10,000
Poor 10,000 20,000 – 15,000
Prepare the expected value table, and advise the management about the choice of product.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 9/2/2019 8:56 PM via UNIVERSIDAD NACIONAL ABIERTA Y A DISTANCIA - UNAD AN: 1950625 ; Sharma, J. K..; Operations Research : Theory and Applications Account: ns145102