DECISION THEORY

INTRODUCTION
📌In many real life problems we have to take decision
under uncertainty.

📌For example, a news boy has to take decision as to how

many newspapers he has to buy from the wholesale agent
each day; a company has to decide whether to go for the
construction of a large plant or a small plant, a milkman
has to decide how many bottles of milk he has to keep in
stock for each day selling and there are many other
problems. 2
INTRODUCTION

📌In these case decision has to be taken under uncertain

future conditions. In these case the use of the probability
distribution of the past data will help the decision maker
to choose the best choice.

3
 Decision making under Uncertainty - Steps

👉What are the possible alternatives to be considered for

taking a decision?

These alternatives are called “states of action”

4
 Decision making under Uncertainty - Steps

👉What are the possible events that may occurs?

It may be possible to identify all the possible events, but the

difficulty is to identify the event which will occur. These
possible events are called the “states of nature”.

5
 Decision making under Uncertainty - Steps

👉What are the possible profits for the different

alternatives under different states of nature?

These profits are called “Pay –Offs”

6
EXAMPLE : 1
A management is faced with the problem of choosing one of the four
products for manufacturing. The potential demand for each product may
turn out to be good, fair, and poor. The estimated profits under these three
states of demand is respect of each product given below:

Products Good Fair Poor

A 50,000 25,000 10,000

B 60,000 30,000 12,000

C 70,000 35,000 5,000

D 80,000 40,000 Loss of 5,000

7
The table shows,

📖the different products are called as alternatives or states of

action;

📖the states of demand are the possible events and they are called the
states of nature.

📖The profits(in Rs.) given in table are called the possible pay-offs.

In general, a pay-off table with ‘m’ states of action and ‘n’ states
of nature is given.
8
In this table. The pay-off corresponding to the act

9
 Decision under different environment

◈ Decision making under conditions of certainty

◈ Decision making under conditions of uncertainty

◈ Decision making under conditions of risk

10
Decision making under conditions of certainty
◈ In this state of nature even if there is certainty it is impossible to
guarantee complete certainty about the future.

◈ Since only one states of nature exists it is easy for decision

maker to choose the best alternative based on pay-offs.

◈ In the second type even though various states of nature exists the
decision maker has no knowledge about the probabilities of
occurrence of these states of nature. 11
Criteria for decision making under uncertainty
Generally the following criteria are considered for decision making under
uncertainty. They are

◈ The Maximax criterion

◈ The Maximin criterion

◈ The Minimax regret criterion

◈ Hurwicz criterion [criterion of realism]

◈ Laplace criterion [criterion of rationality]

12
Example
Suppose a company can produce any number of the products
A,B,C and the payoffs under different states of nature for
demand, good, fair and poor are known

Product Good Fair Poor

A 70,000 40,000 -20,000

B 60,000 65,000 -10,000

C 55,000 45,000 15,000

13
1.Maximax Criterion
It is optimistic rule which maximises the maximum that can be gained under
different products (alternatives). The maximum pay-offs for the products A,B,C
are 70,000;65,000; and 55,000 respectively

Product Good Fair Poor Maximum

Pay-offs
A 70,000 40,000 -20,000 70,000

B 60,000 65,000 -10,000 65,000

C 55,000 45,000 15,000 55,000

14
1.Maximax Criterion
The maximum of these three maximum is 70,000. therefore the decision maker
has to choose the product A according to maximum criterion

Product Good Fair Poor Maximum

Pay-offs
A 70,000 40,000 -20,000 70,000

B 60,000 65,000 -10,000 65,000

C 55,000 45,000 15,000 55,000

15
1.Maximax Criterion

◈ Criterion of OPTIMISM
◈ Maximization of maximum profit

16
2.Maximin Criterion
It is a criterion for the selection of the best among the worst.

In the above table first we choose the worst pay-offs corresponding to

each product.

Product Good Fair Poor Minimum

A 70,000 40,000 -20,000 -20,000

B 60,000 65,000 -10,000 -10,000

C 55,000 45,000 15,000 15,000

17
2.Maximin Criterion
Among these minimum, we select the best pay-off. The decision maker
has to choose the product C according to maximin criterion.

Product Good Fair Poor Minimum

A 70,000 40,000 -20,000 -20,000

B 60,000 65,000 -10,000 -10,000

C 55,000 45,000 15,000 15,000

This is the cautious decision rule based on maximising the minimum

pay-off. 18
2.Maximin Criterion

◈ First suggested by Abraham Wald

◈ Wald Criterion
◈ Criterion of Pessimism
◈ Maximization of minimum pay-offs.

19
3.Minimax Regret Criterion
From the pay-off table let us prepare the regret table.

The regret is the opportunity loss from taking a decision.

Regret Matrix = Max.pay off subtracted from the actual pay off achieved

20
 Product Good Fair Poor

A 70,000 40,000 -20,000

B 60,000 65,000 -10,000

C 55,000 45,000 15,000

GOOD= 70,000-70,000=0 : 70,000-60,000=10,000; 70,000-55,000=15,000

FAIR = 65,000-40,000=25,000; 65,000-65,000=0; 65,000-45,000=20,000

POOR=15,000+20,000=35,000; 15,000+10000=25,000; 15,000-15,000=0

21
Product Good Fair Poor

A 0 25,000 35,000

B 10,000 0 25,000

C 15,000 20,000 0

22
Among these maximum values we choose the minimum. The minimax regret value is
20,000 and hence according to this criterion product C has to be chosen.

Product Good Fair Poor Maximum

regret
A 0 25,000 35,000 35,000

B 10,000 0 25,000 25,000

C 15,000 20,000 0 20,000

Minimax regret value

23
4. Hurwicz Criterion
This is a weighted average of the best and the worst pay-offs of each alternatives.

This may be considered more realistic criterion between optimism and pessimism

This weighted average is defined by

Weighted pay-offs= ∝ (maximum pay-offs) + (1-∝) (minimum pay-offs)

24
Example
Suppose a company can produce any number of the products
A,B,C and the payoffs under different states of nature for
demand, good, fair and poor are known

Product Good Fair Poor

A 70,000 40,000 -20,000

B 60,000 65,000 -10,000

C 55,000 45,000 15,000

25
4. Hurwicz Criterion
If the decision maker is fairly optimistic and chooses a value of
0.7 for their weighted pay-offs for the different products A, B, and C
are
0.7 x 70,000 + 0.3 (-20,000) = 43,000
0.7 x 65,000 + 0.3 (-10,000) = 42,500
0.7 x 55,000 + 0.3 (15,000) = 43,000

The realism criterion suggests that the decision maker can choose
either product A or product C. 26
5. Laplace Criterion
◈ This criterion, also known as the principle of rationality.

◈ Associates equal probabilities for the different states of

nature in the absence of information about these
probabilities.

◈ Therefore the expected payoffs for each pay-off for each

alternative divided by the number of the states of nature.
27
5. Laplace Criterion
For the example, the above payoffs according to laplace criterion for the products A, B, C are

70,000 + 40,000 - 20,000 = 30,000

3

60,000 + 65,000 -10,000 = 38,333

3

55,000 + 45,000 + 15,000 = 38,333

3
28
Therefore to laplace criterion the choices are products B (or products C)
PRACTICE QUESTIONS
1.

EVENTS S1 S2 S3 S4

ACTION

A1 27 12 14 26

A2 45 17 35 20

A3 52 36 29 15
PRACTICE QUESTIONS
2.
EVENTS S1 S2 S3

ACTION

A1 56 8 4

A2 55 52 3

A3 50 44 1