Statistical Decision Theory

Every individual, big or small, rich or poor, educated or uneducated has to take decision almost every day. Some of these decisions are of a routine type which do not involve high states and are consequently trivial in nature. Example A student may decide whether to put on a white or a red shirt while going to the college on a particular day or a house wife may decide to serve lemon uice or pineapple uice to a guest or business e!ecutive may decide whether to go by train or car to meet a potential customer in a nearby town. "owever, in contract to these situations, we have quite often to make decisions which we consider to be important from many points of view and which entails a lot of reasoning and thinking. Example #ecision to buy or not to buy shares of particular company, to accept or not to accept a new recruitment and promotions policy are, by any standard, significant and important decisions and consequently would not be made in a haste or without a detailed analysis of the various pros and cons involved in each situation. Some of the decisions, like buying or not buying the shares of a company or accepting a new ob affect the decision maker only. "e alone has to suffer the consequences of his decision or his family members may also be affected by it. Such decisions are personal decisions. As against these some people have to make decisions which affect other people like consumers of the products, shareholders of the business unit and employees of the organi$ation.

 Such decisions which affect other people in a society require very careful and objective analyses of the different group of persons are differently affected by these and their interests often clashes. Decision Making Science or an Art %he very first doubt that arises in the mind of a common person is that whether decision making could ever be science obeying definite laws which may give high of precisions to the consequences of a particular decision. %he ustification for this doubt is because in decision making we make inferences about unknown rather than known. %here is much that we do not know about all the problems that beset us and that is why most of the time we find ourselves guessing and making sub ective decisions. &ery often it is felt that the choice of the decision is related to the personality of the decision maker and his sub ective assessment of the situation that an unemotional and abstract scientific analysis is in appropriate in this area. 't is also argued that successful men in business have made right decisions without the aid of scientific tools and techniques and that the choice of a decision is largely a matter of intuition based on e!perience. (n a closer e!amination it would become obvious that this feeling is not right and its e!ists because people en oy guessing. (therwise we cannot e!plain the popularity of many games of chance which make guessing synonymous with entertainment. %here can be no shortage of quantitative techniques which can be appropriately and profitably used in such situations and yet are not made rise of our shortage is only of more effective techniques and also our willingness to apply the techniques we already have. Even if it is accepted that there is always a strong human element in decision making in the final stage, process which are amenable to scientific analysis and treatment.

 Efforts are made to evolve a method by which these components could be woven to help the person to make a coherent and consistent decision. 't can, therefore, be stated that decision making in the field of business need not be e!clusive sub ective in character. %here are many areas where it is possible to apply statistical tools and techniques and thereby make decision making more ob ective in nature. #ecision making would, therefore, remain both an art as well as science. Elements in the Decision Making )or all decisions whether routine or comple! there are some common elements. A decision situation arises only when the decision maker has more than one course of action open to him. 'f there is only one alternative, there is nothing to decide. %he first step, therefore, is any decision situation is to find out and list all possible alternatives available. %he list of alternative choices should be as far as possible e!haustive and the list so drawn should provide courses of action which are e!clusive of each other. 't means that out of various courses of action which are e!clusive of each other. 't means that out of various courses of action, if any one is chosen, the others have to be re ected. (ne and only one decision can be choose as it is the best. Another common element in most of decision problems is uncertainty. %his uncertainty is referred to as State of Nature (N) or the State of the World. %he strategy which a decision maker has to choose would depend on the level of uncertainty of the state of nature.

 Statisticians make an effort to reduce the element of uncertainty by trying to assign probabilities to various states of nature on the basis of past records about the problem under study. Measuring Conse uences of !arious Decision 'n order to select a strategy from amongst the various available strategies one has to know the consequences of selecting different strategies. 'n other words one should know the e!tent to which a particular strategy would achieve the ob ective which the decision maker has in his mind. %he problem before the decision maker would be a problem of measurement of the e!tent to which an ob ective is being reali$ed. "ere the decision maker has to face a new type of problem. %here are some ob ectives which are either achieved or not achieved, for e!ample, whether a new product is developed or not developed whether a particular quality is achieved or not achieved. (n the other hand, there are many ob ectives which provide a vast range of attainments. Such ob ectives are profits, cost, sales, production, employee motivation or market goodwill. Some of these ob ectives offer natural way of measuring the degree of achievement in term of say rupees or numbers or some other units. *hereas, there are some other ob ectives like goodwill or employee morals or motivation which cannot be measured in this manner. "owever, when we use+quantitative techniques in decision making these qualitative terms have to be studied in the light of related characteristics. )or e!ample, labour satisfaction may be studied through labour turn over rate or labour productivity etc. 't should be remembered that when our ob ectives can be directly defined in terms of a natural unit, the inferences have to be carefully drawn. ,rofit in terms of rupees is a natural way of measuring profit but does it really reflects the e!tent of

achievement of the ob ectives of higher profits. %he value of rupee differs from unit to unit and individual to individual. %hus, the same rupees have different measures to different persons. 't is the utility which is more relevant measure of achievement of an ob ective rather than the natural unit like a rupee or any other unit of measurement. 'f the ob ectives to be achieved are multiple and are of types which cannot be measured on same scale, the problem becomes more comple!. E!ample, if the ob ectives of the firm are, higher profit, higher productivity and a high quality standard of the product, then there is no single scale of measurement of these ob ectives. "owever, in such comple! situations also, efforts are made to convert these ob ectives into a single utility measure. "ay #ff *hen the value of a consequence is e!pressed directly in terms of gains e!pressed in money, it is called a $pay off. %he consequences of various decisions are given monetary values and when the conditional outcome of the various strategies to possible states of nature is put in the shape of a table, it is called "ay off Table or "ay off Matrix. A pay of matri! takes into account two things a. b. Alternative strategies /or alternatives. which may be denoted by S0, S1, 2S &arious states of nature denoted by 30, 31,23k.

(bviously, the total number of cell in a ,ay off 4atri! would depend on the number of strategies which are available and the various states of nature which are identified. 'f there are 5 strategies and there are 6 states of nature, the total number of cells in the pay off matri! would be 5 ! 6 or 17. A pay off matri! generally assumes the following shape.

State of Nature Strategy

%& ,00 ,10 ,80 ,50 2.. ,0

%' ,01 ,11 ,81 ,51 2.. ,1

%( ,08 ,18 ,88 ,58 2.. ,8

%) ,05 ,15 ,85 ,55 2.. ,5

*++ 2.. 2.. 2.. 2.. 2.. 2..

%k ,0k ,1k ,8k ,5k 2.. ,k

S0 S1 S8 S5 2..2.. S

%here is no rigidity about the rules that state of nature be shown in column and the strategies in rows. %he state of nature can be shown in rows and strategies or actions in columns. 'n pay off matri! ,00 in the pay off the strategy S 0 when the state of nature is 30. Similarly, ,05 is the pay off of strategy S0 when the state of nature is 35. 9ikewise, ,0k will be the pay off strategy S0 when the state of nature is 3k. Example A carpenter is offered 6 tables for :s. 877 /:s. ;7 per table.. "e thinks that he may be able to sell each table for :s. 067 this making a profit of :s. <7 per table /provided he is able to sell them.. ,repare a pay off matri! to depict the decision situation. Solution %he value of pay off matri! can be easily calculated. 'f the state of nature to 37, which means no table can be sold then the entire investment of :s. 877 is a loss. 'f one table is sold for :s. 067, the pay off is :s. 067 = :s. 877 or = :s. 067. 'f two tables are sold the pay off is :s. 877, the pay off would be :s. 877 = 877 or :s. 7. 9ikewise if 8, 5, or 6 tables are sold the pay off would be :s. 067, :s. 877 and :s. 577 respectively.

'f the decision is not to buy, the pay off matri! would show 7 in all states the nature. "ay #ff Matrix
State of Nature Strategy

%, +877 7

%& +067 7

%' 7 7

%( 067 7

%) 877 7

%567 7

S0 = >uy S1 = #o not >uy

)rom the above table it is clear that if the carpenter takes a decision to buy the tables, he will earn of profit of :s. 567 only if he can sell 6 tables at :s. 067 each and he will have no profit /if he sells only 1 tables.. 'f he sells less than two tables then he suffers also. Some times instead of gain we may measure less or cost and in such cases the table is known as 9oss or ?ost %able. 9oss or ?ost is a negative gain. 'n pay off table we select a strategy which gives us ma!imum pay off or minimum cost or loss. Savage has suggested a different measure to assess the consequences of a strategy. According to him if we have selected a strategy and we know our pay off. 3ow we should compare this pay off with the ma!imum pay off which we might have got had we selected another in the state of nature which has occurred. %he ma!imum pay off minus the pay off which we have received is the .egret of the decision makers. )or e!ample, if the pay off of a decision maker under the strategy chosen with the state of nature that has occurred is :s. 677 and if an alternative strategy under this state of nature would have given him :s. @77, then his regret is :s. @77 = :s. 677 or :s. 877. Savage argues that the decision maker should try to minimi$e this regret.

 't is called minimi$ing the regret or loss of opportunity. %he regret or the opportunity loss is simply the difference between the pay off reali$ed and the ma!imum pay off which could have been reali$ed if another strategy was chosen. .egret Table from the above e!ample would be of the following form Sate of %ature Strategy S0 >uy S1 #onAt >uy %, 877 7 %& 067 7 %' 7 7 %( 7 067 %) 7 877 %7 567

)rom the above it is clear that if the carpenter decides to buy the tables his ma!imum regret if the State of 3ature is 3 7 is :s. 877B+ and in case of 30 it is :s. 067B+. After this he has no regret. 'n case he selects strategy 1 /donAt buy. his regret would start when the state of nature would be 38, 35 or 36. 'f the state of 3ature is 37, 30, or 31, he has no regret. Example A fruit dealer buys oranges at the rate of :s. 8 per do$en and sells them at the rate of :s. 6 per do$en. (ranges not sold during the day are treated as stale and thrown away. %he daily sale of oranges in the past has not been less than <@ do$ens and not more than 077 do$ens. ,repare a pay off table and a loss table. Solution %he fruit dealer would not buy less than <@ do$ens and not more than 077 do$ens. So he has three strategies or options open to him. "e buys, <@ do$ens, or << do$ens or 077 do$ens. As such S0 C <@, S1 C << and S8 C 077 do$ens. %hus, 30 C <@, 31 C << and 38 C 077.

 %he e!pected pay off /E,., under various strategies with different states of nature will be,00 C pay of strategy, is S0 and state of nature is 30 ,00 C D<@ ! 6 = <@ ! 8E C :s. 0<; ,01 C pay of strategy, is S0 and state of nature is 31 ,01 C D<@ ! 6 = <@ ! 8E C :s. 0<; ,08 C pay of strategy, is S0 and state of nature is 38 ,08 C D<@ ! 6 = <@ ! 8E C :s. 0<; ,10 C pay of strategy, is S1 and state of nature is 30 ,10 C D<@ ! 6 = <@ ! 8E C :s. 0<; ,11 C pay of strategy, is S1 and state of nature is 31 ,11 C D<< ! 6 = << ! 8E C :s. 0<@ ,18 C pay of strategy, is S1 and state of nature is 38 ,18 C D<< ! 6 = << ! 8E C :s. 0<@ ,80 C pay of strategy, is S8 and state of nature is 30 ,80 C D<@ ! 6 = 077 ! 8E C :s. 0<7 ,81 C pay of strategy, is S8 and state of nature is 31 ,81 C D<< ! 6 = 077 ! 8E C :s. 0<6 ,88 C pay of strategy, is S8 and state of nature is 38 ,88 C D077 ! 6 = 077 ! 8E C :s. 177 Strategy S& /012 State of %ature 30 /<@. 0<; S' /002 0<8 S( /&,,2 0<7

31 /<<. 38 /077.

0<; 0<;

0<@ 0<@

0<6 177

.egret Table Strategy S& S' S( State of %ature 30 /0<;+0<;. /0<;+0<8. /0<;+0<7. 7 8 ; 31 /0<@+0<;. /0<@+0<@. /0<@+0<6. 1 7 8 38 /177+0<;. /177+0<@. /177+177. 5 1 7 Decision Types #ecision problems can be classified in many ways. "owever, of all the classifications, the one which is based on the quantum of information available is considered to be most useful and rational. (n this basis, decision problems can be classified in the following five categories. /i2 Decision making under certainty 'n certain problems the information available to the decision maker is almost complete so that he knows all the facts about the states of nature /risk. and also knows which state of nature /probability. would occur. 'n such situation the problem of decision making is simple. %he decision maker has to choose the strategy which will give him ma!imum pay off in terms of utility under the state of nature which he knows will occur. /ii2 Decision making under risk %his type of problem arises when the state of nature is unknown, but on the basis of ob ective or empirical evidence it is possible to assign probabilities to various states of nature.

 'n such cases the pay off matri! is of great help and an optimal decision can be arrived at by assigning probabilities to various states of nature. /iii2 Decision making under uncertainty

A decision process is said to be under condition of uncertainty when in states of nature are unknown and no ob ective information is available about their probabilities of occurrence. 'n such cases there is no historical data or no relative frequency which could indicate the probability of occurrence of a particular state of nature. Such situation arises when a new product is introduced in the market or a new plant is set up. (f course even in such cases some market surveys are conducted and relevant information gathered but generally it is not sufficient to indicate a probability figure for the occurrence of a particular state of nature. /iv2 Decision making under partial information

%his is a situation some where between the condition of risk and the condition of uncertainty. 'n the case of condition of risk the probability of the occurrence of various states of nature is known on the basis of past e!perience and in condition of uncertainty, there is no such data available. "owever, there might be many situations where thee is partial information available of the data. 'f is so, decision making is said to done on the basis of partial information.

/v2

Decision making under conflicts A condition of conflict is supposed to be e!it when instead of state of nature we are dealing with rational opponent. "ere the decision

maker has to choice a strategy which takes into account the action and counter action of his opponents. 4arket place, brand competition military weapons etc. are problems which come under this category. %he choice of strategy in such situationBconditions is done on the basis of game theory 3here a decision maker anticipates the action of his opponent and then determines his o3n strategy. 't is like paying a game of chess. Choice of a Decision Criteria 'n order to select a strategy from amongst the many in different types of decision situations, it is necessary that the selected strategy is one which is most appropriate for achieving the ob ective in mind of the decision maker. %he nature of decision criteria would depend on the types of the decision situation. 4nder condition of certainty Fnder the condition of certainty there is a pay off for each strategy. %he pay off measured as utility, i.e., profit represents the degree of achievement of the ob ective, hence the largest pay off is chosen and the corresponding strategy is selected. 'f, however, the measure is the cost, then the strategy with the lowest cost is picked up. 4nder condition of risk Fnder the condition of risk there would be more than one state of nature but the probabilities of their occurrence are known on the basis of their past e!perience. 'n this situation each strategy will have as many pay off as the states of nature. %o pick up the correct strategy we will have to transform all possible pay offs of a strategy into a single figure, on

the basis of their probabilities of the states of nature and the e!pected pay off. %he strategy which gives ma!imum pay off is selected in such cases. 4nder condition of uncertainty Fnder the condition of uncertainty, since we do not know the probability of the occurrence of various states of nature, the problem becomes more comple! and the personality of the decision maker plays an important role in the selection of the strategy, thus the decision taken under uncertainty are necessarily sub ective. 5o3ever6 the analyst has devised some decision rules to impart some objectivity to the subjective decisions . %he following decisions choices reflect the attitude of the decision maker. /i2 7ald8s Maxim in Decision criterion *aldAs ma!im in decision criterion tells us that the decision makers should specify first the worst possible outcome of each strategy and accept a strategy that gives the best out of the worst outcomes. %o illustrate the e!ample let us consider a hypothetical pay off matri! as "ay off Matrix States of %ature Strategy S0,lant G 4achinery S1 Equity S8Hovernment >onds S5 :eal Estate %& 17 06 0; 6 %' 01 0; @ 01 %( ; 5 ; 8 %) 6 +1 +0 1

't is assumed that an investor conceives of four strategic investment pro ects S0, S1, S8 and S5 /investment in plant and machinery, equity, real estate and government bonds respectively. under four different states of nature of the economy, 3 0, 31, 38

and 35 /high growth, low growth, stagnation and recession respectively.. %o apply ma!imin criterion, the decision maker needs to find out the worst /minimum. outcome of each strategy. %his can be done by reading the pay off table row wise. %he ma!imin column presents the worst outcome of each strategy. %he best or the highest outcome of the worst outcome is 6 of strategy. Hoing by the ma!imin criterion, the decision maker would accept strategy S0. 'f the ma!imin rules are closely looked, it implies a pessimistic approach to investment decision making. 9t gives a conservative decision rule for risk avoidance. /ii2 Maximax Criteria According to this criterion, if the decision maker is an optimist by nature he would always think that the state of nature would be the best from his point of view. "e would find out the e!pected pay off of all the strategies and will pick up the strategy which gives the ma!imum pay off out of ma!imum pay offs of all the strategies. "e always thinks that the states of nature would be favourable to him and his eyes are on the ma!imum possible pay off of all the strategies. /iii2 Minimax .egret Criteria 4inima! regret criterion is another decision rule under uncertainty. %his criteria suggests that the decision maker should select a strategy that minimi$e the maximum regret of a wrong decision.

Example Suppose an investor has three strategies for investment, S0, S1 and S8, giving returns of :s. 07,777B+ :s. @,777B+ and :s. ;,777B+ respectively. 'f the investor opts for strategy S0, he gets the ma!imum possible return, he has no regret. >ut, if he opts for S 1, by way of incorrect decision, then his regret on opportunity cost would be :s. 07,777B+ = :s. @,777B+ C :s. 1,777B+. 9ikewise if he opts for S8, his regret equals :s. 07,777B+ + :s. ;,777B+ C :s. 5,777B+. Hoing by the minima! regret criterion, the investor should opt for strategy S1 because it minimi$e the regrets. Suppose we have the following pay off table, we can construct the regret table. %he method is simple. Select a column /the state of nature., find the ma!imum pay off and subtract it from the pay off of all strategies. %his process gives the pay off column.

Strategy S0 S1 S8 S5

States of %ature %& %' %( %) 17 01 ; 6 06 07 5 +1 0; @ ; +0 6 01 8 1

.egret Matrix %& %' %( %) 7 7 7 7 6 1 1 8 5 5 7 5 06 7 8 8

Max:Min .egret 7 6 5 06

)or e!ample, under column 30, strategy S0 has the ma!imum pay off /17.. 't means that if S 0 is chosen under the states of nature 30, the regret is $ero. 'n the ne!t strategy S1, the pay off is 06 and the regret is equal to 6. >y repeating this process for all strategies and all the states of nature we get regret matri!. )rom the regret

matri! we can find Ima!imin regretA by listing the ma!imum regret for each strategy. %he regret table shows that the ma!imum regret is minimum /06. in case of strategy S5. %herefore, strategy S5 strategy should be selected for investment. /iv2 ;aplace Decision criterion %he laplace criterion was >ayesian rule to calculate the Iexpected valueA of each strategy. According to the >ayesian rule, where meaningful estimate of probabilities is not available, the outcome of each strategy under each state of nature must be assigned the same probability and the sum of probabilities of out come of each strategy must add up to one. )or this reason, the 9aplace criterion is also called I>ayesian criterionA. >y assuming e ual probability for all events, the environment of uncertainty is converted into an environment of risk %his decision rule avoids the problem that arises due to sub ectivity in assuming a probability of pay off. Example A news paper vender buys a newly started local paper for 6 n.p. and sells it at the rate of 07 n.p. %he unfold paper do not have any value. %he vendor knows that he cannot sell more than 17 papers in a day and the minimum sale would not be less than 05. "ow many papers should he buy. Solution %he strategies open to the newspaper vender are to buy 05, 06, 0;, 0J, 0@, 0< or 17 papers.

 %he states of nature are the number of newspaper he can sell. %he states of nature would be 05, 06, 0;, 0J, 0@, 0< or 17. ,rofit of vendor wills as follows if he buys only 05 profit 6 n.p. on each paper J7 n.p. 'f he sells 06 papers he earn a profit of J6 n.p. 9ikewise, the profit of other class can be calculated. Strategy S&) S&- S&< S&= S&1 S&0 S', States of %ature 305 J7 ;6 ;7 66 67 56 57 306 J7 J6 J7 ;6 ;7 66 67 30; J7 J6 @7 J6 J7 ;6 ;7 30J J7 J6 @7 @6 @7 J6 J7 30@ J7 J6 @7 @6 <7 @6 @7 30< J7 J6 @7 @6 <7 <6 <7 317 J7 J6 @7 @6 <7 <6 077 Note > 9n this problem no probability values are assigned to various states of nature6 hence it is a situation of uncertainty+ 'n a situation of uncertainty, the decision makers personality would be reflected in the decision he takes. )or instance, if he is pessimistic he would follow the criteria Maximin. 'f he is an optimistic he would follow the criteria ma!ima!. "e may also think in terms of minimi$ing the ma!imin regret or try to estimate the opportunity loss. )or this we would have to prepare a regret table.

0.

4a!imin #ecision

1.

4a!ima! #ecision

8.

4inimi$ing ?riterion

the

4a!imum

:egret

>uy 05 papers. "e will ma!imi$e the minimum pay off. %he largest minimum pay off is J7, so he will buy 05 papers. >uy 17 papers. %he highest ma!imum pay off is 077. "e will buy 17 papers as it ma!imi$es the pay off. )or studying this it is necessary to construct or prepare a regret table.

.egret Table Strategy States of %ature 305 306 30; 30J 30@ 30< 317 S&) 7 6 07 06 17 16 87 S&6 7 6 07 06 17 16 S&< 07 6 7 6 07 06 17 S&= 06 07 6 7 6 07 06 S&1 17 06 07 6 7 6 07 S&0 16 17 06 07 6 7 6 S', 87 16 17 06 07 6 7

3ow if the state of nature is 305 and S05 the regret would be 7 as he loses nothing but if the strategy is S05 and state of nature is 306 his regret would be 6, because the ma!imum pay off with 306 to J6 and he would reali$e only J7. 9ikewise with S05 and 30; his regret would be 07 as the ma!imum pay off under 30; is @7. 4nder Minimax Criteria "e will buy 0J news paper. %here he is minimi$ing the ma!imum regret. %he ma!imum regret with S0J C 06. 'n all the other strategies the ma!imum regret is more than 06.

 All the above decision have been taken under conditions of uncertainty. %he decision maker did not have any probability values for various states of nature. 'f the decision maker on the basis of past e!perience could assign some probabilities to the various states of nature.

State of %ature 305 306 30; 30J 30@ 30< 317

"robabilities 7.7@ 7.11 7.17 7.05 7.05 7.01 7.07 0.77

*ith the above information the decision maker is in a better position to take a rational decision. %he e!pected pay off table would be. "ay off Table Strategy States of %ature 305 306 30; 30J 30@ 30< " 77.7 @ 77.1 1 77.1 7 77.0 5 77.0 5 77.0 S&) 76.; 7 06.0 5 05.7 7 7<.@ 7 7<.@ 7 7@.5 S&76.1 7 0;.6 7 06.7 7 07.6 7 07.6 7 7<.6 S&< 75.7 @ 06.5 7 0;.7 7 00.1 7 00.1 7 7<.; S&= 75.5 7 05.8 7 06.7 7 00.< 7 00.< 7 07.1 S&1 75.7 7 08.1 7 05.7 7 00.1 7 01.; 7 07.@ S&0 78.; 7 01.0 7 08.7 7 07.6 7 00.< 7 00.5 S', 78.17 00.77 01.77 7<.@7 00.17 07.@7

317

1 77.0 7 70.7 7

7 7J.7 7 J7.7 7

7 7J.6 7 J5.1 7

7 7@.7 7 J;.1 7

7 7@.6 7 J;.1 7

7 7<.7 7 J5.@ 7

7 7<.6 7 J1.7 7

07.77 ;@.77

(n the basis of the above e!pected pay off table, the highest e!pected pay off of J;.17 is associated with two strategies S 0; and S0J. %herefore, the vendor should buy either 0; or 0J papers. 'f the decision is to be taken on the basis of regret table, the result would also be the same. .egret Table Strategy States of %ature 305 306 30; 30J 30@ 30< 317 " 77.7 @ 77.1 1 77.1 7 77.0 5 77.0 5 77.0 1 77.0 7 70.7 7 S&) 77.7 7 70.0 7 71.7 7 71.0 7 71.@ 7 78.7 7 78.7 7 05.7 7 S&77.5 7 77.7 7 70.7 7 70.5 7 71.0 7 71.5 7 71.6 7 7<.@ 7 S&< 77.@ 7 70.0 7 77.7 7 77.J 7 70.5 7 70.@ 7 71.7 7 7J.@ 7 S&= 70.1 7 71.1 7 70.0 7 77.7 7 77.J 7 70.1 7 70.6 7 7J.@ 7 S&1 70.; 7 78.8 7 71.7 7 77.J 7 77.7 7 77.; 7 70.7 7 7<.1 7 S&0 71.7 7 75.5 7 78.7 7 70.5 7 77.J 7 77.7 7 77.6 7 01.7 7 S', 71.57 76.67 75.77 71.07 70.57 77.;7 77.77 0;.77

)rom the above table it is clear that the minimum regret of J.@ is associated with S0; and S0J. As such he should buy either 0; or 0J papers. ;aplace Decision

 'n 9aplace decision theory, the decision maker /in the absence of any past experience2 3ould assign e ual probabilities to all states of nature. Since in this case there are J states of nature, hence the probability of each state of nature would be 0BJ. %he e!pected pay off under laplace decision would be obtained by multiplying each pay off with 0BJ. %he pay off matri! would there be as follows;aplace "ay #ff Table Strategy States of %ature 305 306 30; 30J 30@ 30< 317 " 0BJ 0BJ 0BJ 0BJ 0BJ 0BJ 0BJ S&) 07.7 7 07.7 7 07.7 7 07.7 7 07.7 7 07.7 7 07.7 7 J7.7 7 S&7<.1 < 07.J 0 07.J 0 07.J 0 07.J 0 07.J 0 07.J 0 J8.6 6 S&< 7@.6 J 07.7 7 00.5 1 00.5 1 00.5 1 00.5 1 00.5 1 J6.; J S&= 7J.@ 6 7<.1 @ 07.J 0 01.0 5 01.0 5 01.0 5 01.0 5 J;.5 7 S&1 7J.0 5 7@.6 J 07.7 7 00.5 1 01.@ 6 01.@ 6 01.@ 6 J6.; @ S&0 7;.5 1 7J.@ 6 7<.1 @ 07.J 0 01.0 5 08.6 J 08.6 J J8.6 5 S', 76.J0 7J.05 07.77 00.51 01.@6 05.@6 05.1@ ;<.<J

According to the 9aplace pay off matri! the ma!imum pay off is J;.57 which is associated with S0J. Decision:Tree Method 4nder the conditions of risk6 the decision makers often confront a situation is which they visuali$e several option available to them each leading to different states of nature.

 %here are two specific problems in this kind of decision making. ?irst, the decision makers are required to make a choice /or series of choices. from the alternative investment avenues to them. %hey are not supposed to leave the matter undecided. Second, the decision makers know for sure that all the decisions will yield a positive outcome, but they cannot tell in advance the exact outcome of a decision. %hey might be knowing that a particular decision will yield a higher return than another but they do not know for sure high or low the outcome will be. %he question that decision+makers face under these conditions is how to find the most profitable or gainful solution. %he method that is used to find an acceptable solution under these conditions is called a Tree:Decision. A tree+decision is a graphical device to map all possible managerial decisions in a sequence and their e!pected outcomes under different states of economy. Since all possible strategic decisions and their possible outcomes are arranged graphically in the form of branches of a tree, the technique is called decision:tree. %he decision+tree presents the entire decision options and possible outcomes in the form of a diagram and thereby guides the decision meters to a rational decision. Decision Tree for 9nvestment Decision

 Suppose an investor visuali$es two viable pro ects A and >. %he pro ect A costs :s. 677 crore and ,ro ect > cost :s. 57 crore. %he prospective yield /cash flow. of the two pro ects depend upon the states of the economy, i.e. whether the economy has a high or a low growth. %he probable growth rates of the economy determine the demand prospects for the product of each pro ect. #emand may be high, medium or low, depending on the market conditions and consumerAs perception and preference. %he investor has the information about the e!pected cash flows under the different states of economy and the demand prospects. %he decision shows that the process of decision+making begins at Idecision+pointA. %he decision maker has to make a strategic choice between the two pro ects A and >. ?olumn /1. shows that the probabilities of the Istates of economyA. %he economy may grow at a higher or at a low growth rate. %he high growth has a probability of say 7.; and the low growth has a lower probability of 7.5.

 Fnder both these growth probabilities, the prospect for the product demand again has three probabilities = high, medium and low = under both high and low growth of the economy. 't should also be noted that the probability distribution in respect of demand prospects = high, medium and low = under high growth add up to 7.; and in case of low growth they add up to 7.5. 9et us now suppose that the investor has the information on the present value of cash flow under each probability as presented in col. /6.. 3ow when the present value of cash flow is multiplied by the corresponding probability in col. /5., it gives the Ie!pected valueA of the present value of cash flows. ?ol. /;. gives the e!pected value of the two pro ects under all the stipulated conditions. %he investor has now the full information for decision making. 9nvestment decision %he investor can easily find out the net e!pected value of each pro ect and decide in favour of the pro ect having a higher net e!pected value. otal Expected Value : Rs. 580 Crore

Project ! :

- Less Project Cost : Rs. 500 Crore Net expected Value = Rs. 80 Crore

otal Expected Value : Rs. #"0 Crore - Less Project Cost : Rs. #00 Crore Project $ : Net expected Value = Rs. "0 Crore
According to this calculation, the net e!pected value of pro ect A is higher than the pro ect >. %herefore, a rational investor would decide to invest in pro ect A and not in ,ro ect >. Example %he ?oca+?ola associates deals with instant soft drink. %hey have two courses of action for selling their product in the market /a. :egional distribution through distributors and /b. #irect selling.

%he prior probabilities of high penetration and low penetration of regional distribution channel are 7.J and 7.8 respectively. %he prior probabilities of high penetration and low penetration of direct selling channel are 7.; and 7.5 respectively. %he pay off of high and low penetration of regional distribution channel are :s. 67 lakh and :s. 07 lakh respectively. %he pay off of high and low penetration of regional distribution channel are :s. 67 lakh and :s. 07 lakh respectively. %he pay off of high and low penetration of direct selling channel are :s. 87 lakh and :s. 6 lakh respectively. #raw the decision tree and determine the best selling channel i.e. strategy. Solution

%he total e!pected monetary value by regional distribution 7.J K 67 L 7.87 K 07 C :s. 8@ lakh %he e!pected monetary value by direct selling

 7.; K 87 L 7.5 K 6 C :s. 17 9akh Since the e!pected monetary value is more i.e. regional distribution in more than direct selling he is should opt for regional distribution. ;imitation of Distribution Theory %he decision theories associated with conditions of risk and conditions of uncertainty have many limitations. %heories under conditions of risk and uncertainty e!ist because thee are limitations in the relative frequency approach on the basis of which probabilities are assigned to various states of nature and also because the e!pected pay off matri! does not always provide infallible information for a decision situation.

Decision Tree Method

Dra3ing a Decision Tree Mou start a #ecision %ree with a decision that you need to make. #raw a small square to represent this towards the left of a large piece of paper. )rom this bo! draw out lines towards the right for each possible solution, and write that solution along the line. Neep the lines apart as far as possible so that you can e!pand your thoughts. At the end of each line, consider the results. 'f the result of taking that decision is uncertain, draw a small circle. 'f the result is another decision that you need to make, draw another square. Squares represent decisions, and circles represent uncertain outcomes. *rite the decision or factor above the square or circle. 'f you have completed the solution at the end of the line, ust leave it blank. Starting from the new decision squares on your diagram, draw out lines representing the options that you could select. )rom the circles draw lines representing possible outcomes. Again make a brief note on the line saying what it means. Neep on doing this until you have drawn out as many of the possible outcomes and decisions as you can see leading on from the original decisions. An e!ample of the sort of thing you will end up with is shown in )igure 0-

(nce you have done this, review your tree diagram. ?hallenge each square and circle to see if there are any solutions or outcomes you have not considered. 'f there are, draw them in. 'f necessary, redraft your tree if parts of it are too congested or untidy. Mou should now have a good understanding of the range of possible outcomes of your decisions. Evaluating @our Decision Tree 3ow you are ready to evaluate the decision tree. %his is where you can work out which option has the greatest worth to you. Start by assigning a cash value or score to

each possible outcome. Estimate how much you think it would be worth to you if that outcome came about. 3e!t look at each circle /representing an uncertainty point. and estimate the probability of each outcome. 'f you use percentages, the total must come to 077O at each circle. 'f you use fractions, these must add up to 0. 'f you have data on past events you may be able to make rigorous estimates of the probabilities. (therwise write down your best guess. %his will give you a tree like the one shown in )igure 1-

Calculating Tree !alues (nce you have worked out the value of the outcomes, and have assessed the probability of the outcomes of uncertainty, it is time to start calculating the values that will help you make your decision. Start on the right hand side of the decision tree, and work back towards the left. As you complete a set of calculations on a node /decision square or uncertainty circle., all you need to do is to record the result. Mou can ignore all the calculations that lead to that result from then on. Calculating the !alue of 4ncertain #utcome %odes *here you are calculating the value of uncertain outcomes /circles on the diagram., do this by multiplying the value of the outcomes by their probability. %he total for that node of the tree is the total of these values. 'n the e!ample in )igure 1, the value for Pnew product, thorough developmentP is7.5 /probability good outcome. ! Q0,777,777 /value. C 7.5 /probability moderate outcome. ! Q67,777 /value. C 7.1 /probability poor outcome. ! Q1,777 /value. C L Q577,777 Q17,777 Q577 A)',6),,

)igure 8 shows the calculation of uncertain outcome nodes-

3ote that the values calculated for each node are shown in the bo!es.

Calculating the !alue of Decision %odes *hen you are evaluating a decision node, write down the cost of each option along each decision line. %hen subtract the cost from the outcome value that you have already calculated. %his will give you a value that represents the benefit of that decision. 3ote that amounts already spent do not count for this analysis = these are Psunk costsP and /despite emotional

counter+arguments. should not be factored into the decision. *hen you have calculated these decision benefits, choose the option that has the largest benefit, and take that as the decision made. %his is the value of that decision node. )igure 5 shows this calculation of decision nodes in our e!ample-

'n this e!ample, the benefit we previously calculated for Pnew product, thorough developmentP was Q517,577. *e

estimate the future cost of this approach as Q067,777. %his gives a net benefit of Q1J7,577. %he net benefit of Pnew product, rapid developmentP was Q80,577. (n this branch we therefore choose the most valuable option, Pnew product, thorough developmentP, and allocate this value to the decision node. .esult >y applying this technique we can see that the best option is to develop a new product. 't is worth much more to us to take our time and get the product right, than to rush the product to market. 't is better ust to improve our e!isting products than to botch a new product, even though it costs us less. Bey "oints> #ecision trees provide an effective method of #ecision 4aking because they

?learly lay out the problem so that all options can be challenged. Allow us to analy$e fully the possible consequences of a decision. ,rovide a framework to quantify the values of outcomes and the probabilities of achieving them. "elp us to make the best decisions on the basis of e!isting information and best guesses.

As with all #ecision 4aking methods, decision tree analysis should be used in con unction with common sense = decision trees are ust one important part of your #ecision 4aking tool kit.