Statistical Decision Theory

• “It is concerned with the selection of an

optimal course of action from among several
alternatives where the outcome associated
with an action is uncertain”
 Decision making is an everyday process in life. It is
a major role of manager too. Decisions may be
classified into two categories tactically and
strategically.
• Tactical decisions are those which effects the
business in the short run.
• Strategic decisions are those which have far
reaching effect on the course of business.
There are two basic elements of a decision situation.
1. Alternative / strategies.
2. States of Nature.
 Decision making constitutes one of the highest form of
human activity. The problem under study may be
represented by a model in terms of the following elements:
1. The decision maker: a person who is responsible for making
decisions.
2. The acts: are alternative course of action or strategies that
are available for decision making.
3. Events: are the occurrences that effect the achievement of
the objectives( also known as states of nature)
4. Payoff table: it represents the profits/ loss of a problem.
5. Opportunity loss table: it represents the cost or loss
incurred because of failure to take best possible action. It is
the numerical difference between the optimal outcome and
the actual outcome for a given decision.
 Steps in decision making:
• Identify the alternatives.
• Identify the possible outcomes i.e states of
nature.
• Tabulate the payoff function.
• Construct the opportunity loss table
(difference between the highest possible
outcome & actual outcome).
• Select the optimum decision criterion, which
results the largest payoff.
States of A1 A2
nature

S1 a11 a12

S2 a21 a22
Types of Decision making situations:
•Decision making under certainty.
•Decision making under uncertainty.
•Decision making under risk.
•Decision making under conflict.
Decision making under certainty:
In this case the decision maker knows with certainty the
consequences of every alternatives of decision choice.
The decision maker will choose that strategy which will
yield the maximum profit.
DECISION MAKING UNDER UNCERTAINTY:
When the decision maker have multiple choices but has no
means to arrive at probability value to the likelihood of
occurrence of these states of nature, the problem is a
decision problem under uncertainty.
Following are choices available before the decision maker
in situation of uncertainty.
1. Maximax Criterion.
2. Minimax Criterion.
3. Maximin Criterion.
4. Hurwicz Alpha Criterion.
5. Laplace Criteria
1. MAXIMAX CRITERION: ( CRITERION OF OPTIMISM)
The term maxi max is an abbreviation of the phrase
maximum of maximum.
Suppose for each action there are three possible pay off
corresponding to three states of natures, as given below:

States of Decisions
Maximum of Maximum =
nature
A1 A2 A3 220
A1 strategy to be selected

S1 220 180 100

S2 160 190 180

S3 140 170 200

Maximum 220 190 200

2. MAXIMIN DECISION CRITERION: (CRITERION OF
PESSIMISM)
It is an abbreviation of phrase maximum of minimum .
Ex2: A1 A2 A3 A4
P 8 -4 14 13
Q 4 10 6 8
R 14 12 0 6
S -6 7 4 -3
Minimum -6 -4 0 -3

Maximum of Minimum = 0
A3 strategy to be selected
3. THE MINIMAX LOSS (REGRET)CRITERION:
It is an abbreviation of phrase minimum of
maximum.
• Construct a regret table for every state of nature.
• Select max payoff in each row (states of nature)
& subtract all elements from the maximum value
in the respective row.
• From each alternative (decisions) select the
maximum value.
• Select the alternative with the smallest value in
decision column.
 States of A1 A2 A3
nature/Altern
atives
S1 7,00,000 5,00,000 3,00,000

S2 3,00,000 4,50,000 3,00,000

S3 1,50,000 0 3,00,000

Ex3:
Which strategy should the company choose
according to:
1.Maximin Criterion.
2.Maximax Criterion.
3.Minimax regret Criterion.
PAYOFF TABLE OPPORTUNITY LOSS TABLE
A1 A2 A3 A1 A2 A3

S1 7,00,000 5,00,000 3,00,000 S1 0 2,00,000 4,00,000

S2 3,00,000 4,50,000 3,00,000 S2 1,50,000 0 1,50,000

S3 1,50,000 0 3,00,000 S3 1,50,000 3,00,000 0

A1 A2 A3 Minimum of
S1 0 2,00,000 4,00,000 Maximum
= 1,50,000
S2 1,50,000 0 1,50,000 A1 strategy to
be selected
S3 1,50,000 3,00,000 0
Maximum 1,50,000 3,00,000 4,00,000
 4. LAPLACE CRITERION:
As the decision maker has no information about the
probability of occurrence of various events, the decision
maker makes a simple assumption that each probability is
equally likely.

States Decisions A1: (20+25+30)/3 =25

of
nature A1 A2 A3 A2: 12+15+20 /3= 15.67
A3: 25+30+22/3 = 25.67
S1 20 12 25
Since the expected payoff
S2 25 15 30
for alternative A3 is
higher, it is the one that
S3 30 20 22
should be selected.
5. HURWICZ ALPHA CRITERION:
In this method, the decision maker’s degree of optimism is
represented by  and the coefficient of the optimism  varies
between o and 1.When  =0, there is total pessimism and when
 =1, there is total optimism.
• Find the largest & smallest payoff from each
alternative.
• Assign weights for largest () & smallest (1- )
• Compute the expected value.
• Choose the alternative with highest expected
value.
From the following information available select the
alternatives if the person applies the Hurwicz Alpha
Criterion.

States Decisions Alternatives Largest Smallest Expected

of A1 A2 A3 Value
nature A1 30 20 30*.6+20*.4 =
26
S1 A2 20 12 20*.6+12*.4 =
20 12 25 16.8
S2 25 15 30 A3 30 22 30*.6+22*.4 =
26.8
S3 30 20 22
Given = 0.6
Since A3 has highest payoff, it
should be selected.
DECISION MAKING UNDER RISK: THE
BAYES’ PRINCIPLE
In this situation decision maker has some knowledge
or experience which will enable him to assign
probability to the occurrence of each states of nature.
The objective is to optimize the expected profit or to
minimize the opportunity loss.
Under this situation the methods used are:
•Expected Monetary Value Criterion (EMV)
•Expected opportunity loss criterion (EOL)
•Expected value of perfect information Criterion
(EVPI)
EXPECTED MONETARY VALUE CRITERION
(EMV):
Probabilities are assigned to various states of nature. Value of
each event is multiplied by its probabilities and product is
summed up. The resulting number is the EMV of the act.

States Decisions States

of of Decisions
A1 A2 Prob
nature nature A1 A2

S1 30 20 0.6
S1 18 12
S2 35 30 0.4
S2 14 12
PAYOFF TABLE 32 24

EMV TABLE
 EXPECTED OPPORTUNITY LOSS
CRITERION (EOL):
The difference between the highest and actual payoff is
known as opportunity loss. under this strategy with minimum
opportunity loss is chosen.
States Decisions States Decisions
P
of of
A A1 A2 Prob A1 A2 Prob
nature nature O
Y
O L
F S1 0 10 0.6
F S1 30 20 0.6
S2 35 30 0.4 S2 0 5 0.4

States States
of Decisions of Decisions
E nature A1 nature A1 A2
M A2 E
O
V
L
S1 18 12 S1 0 6
S2 14 12 S2 0 2
EXPECTED VALUE OF PERFECT
INFORMATION CRITERION (EVPI):
It is the average return in the long run, if we have perfect
information before the decision is to be made.

EVPI = EPPI – EMV (max)

EPPI = Expected payoff with perfect information
EPPI = Highest payoff for each event * probability for
corresponding event.
Calculate EMV and EVPI from the following pay off
of three acts and three states of nature.
PAYOFF TABLE EMV TABLE
States Decisions States of Decisions
of nature
A1 A2 A3
nature A1 A2 A3

S1 15 12.5 11
S1 0.5 30 25 22
S2 8 14 8
S2 0.4 20 35 20
S3 4 3 3.5
S3 0.1 40 30 35
SUM 27 29.5 22.5

EPPI = (30 * .5) + (35 * .4) + (40* .1) = 33

EMV: 29.5
EPPI: 33
EVPI: 3.5
Ex6: A grocery store with a bakery department is faced with the problems of
how many cakes to buy in order to meet the day’s demand. The grocer
prefers not to sell day old goods in competition with fresh products. Also if a
customer desires a cake and all of them have been sold, the disappointed
customer will buy else where. Thus he has collected information on the past
sales or a selected 100 day period as given below

Sales per day No. of days Probability

25 10 .10
26 30 .30
27 50 .50
28 10 .10
100 1.00

Construct the payoff table and the opportunity loss table. What is the
optimal number of cakes that should be bought each day. A cake costs
Rs. .80 and sells for Re. 1.
Constructing a pay Off table:
Marginal Profit = Selling Price – Cost Price

Marginal loss = Cost Price of unsold items

Conditional Profit = ( Marginal Profit X Sold

items) - (Marginal loss X Unsold items)
 PURCHASE BY SHOP KEEPER

25 26 27 28

25 (0.1) (25 *0.20)= (25*.20)- (25*.0.2)- (25*0.2)-2.4

5 0.80 =4.2 1.60 = 3.4 =2.6

DEMAND BY 5 (26*0.2) = 4.4 3.6

THE
26(0.3)
5.2
CUSTOMER
27(0.5) 5 5.2 5.4 4.6

28(0.1) 5 5.2 5.4 5.6

 PURCHASE BY SHOP KEEPER

25 26 27 28

25(0.1) 5 (25 4.20 ( 25 3.4 2.6

X .20) X .20- .80
)
DEMAND BY THE 26(0.3) 5 5.2 (26 4.4 3.6
CUSTOMER X .20)

27(0.5) 5 5.2 5.4 4.6

28(0.1) 5 5.2 5.4 5.6

 CALCULATION OF EMV
PURCHASE BY SHOP KEEPER
25 26 27 28

25(0.1) 0.5 .42 .34 .26

DEMAND BY THE 1.5 1.56 1.32 1.08

26(0.3)
CUSTOMER

27(0.5) 2.5 2.6 2.7 2.3

28(0.1) .5 .52 .54 .56

TOTAL 5 5.1 4.9 4.2

Ex 7: . News paper boy has following probability of selling a magazine

No. of copies sold Probability

10 .10
11 .15
12 .20
13 .25
14 .30

Cost of a copy is 30 paise and sale price is 50 paise. He cannot

return unsold copies. How many copies should he order?
DECISION MAKING UNDER CONFLICT OR
COMPETITION:
It is a situation when two or more person
desires a same thing and they compete
with each other. In such situation we say
that their interest is conflicting.(Game
Theory)
 DECISION TREE:
•Decision trees are excellent tools for
helping you to choose between several
courses of action.
•They provide a highly effective structure
within which you can lay out the options and
investigate the possible outcomes of
choosing those options.
 STEPS IN DECISION TREE ANALYSIS
1. Identify the decision points and alternatives at each point
systematically.
2. At each decision point, determine the probability and the
associated payoff with each course of action
3. Compute the expected payoff (EMV) for each course of
action. Start from the extreme right and move towards the
left.
4. Choose the course of action that yields the best payoff for
each of the decisions. Proceed backward to the next stage
of decision points.
5. Repeat the above steps till the first decision point is
reached.
6. Identify the best course of action to be adopted from the
beginning to the end, under the various possible outcomes
as a whole
 Decision Tree

Carry Not
Umbrella(A1 Carry(A2)
)

Rain(E1) Safe and Dry Wet

No rain(E2) Burden Relaxed

 rain E1 SAFE & DRY
A1
No rain E2 BURDEN
Carry
umbrella
1

Not to Carry
umbrella

rain WET
E1
A2

No rain E2 RELAXED
 PURCHASE BY SHOP KEEPER
25 26 27 28

25 0.5 .42 .34 .26

(0.1)
DEMAND BY THE 1.5 1.56 1.32 1.08
26(0.3)
CUSTOMER

27(0.5) 2.5 2.6 2.7 2.3

28(0.1) .5 .52 .54 .56

TOTAL 5 5.1 4.9 4.2

 STATES OF NATURE

DECISIONS