c
 




 

 
 
  c 
 
CONTENTS
m Decision Theory ± Process

m One Stage Decision Theory

m Pay-off & Regret Matrix

m Decision Rules

m Bayesian Rule

m Multi Stage Decision Theory

DECISION THEORY
´ A decision is defined as the selection by the decision maker of an
act, considered to be best according to some standard, from among
the available options.

´ Decision theory is also called the decision analysis.

´ It is used to determine optimal strategies where a decision maker

is faced with several decision alternatives and an uncertain, or
risky, pattern of future events.
DECISION MAKING PROCESS
´ Identification of various possible outcomes, called ë
ë

  ë¢
ë, for the decision problem . The events are beyond the
control of the decision-maker.
´ Identification of all the 6ë
6  ë, or the strategies that
are available to decision-maker. The decision maker has control
over choice of these.
´ Determination of the pay-off function which describes the
consequences resulting from the different combinations of the acts
and events. The pay-offs may be designated as Vij¶s - the pay off
resulting from the ith event jth strategy
´ Choosing from among the various alternatives on the basis of some
criterion, which may involve the information given in step 3rd only
or which may require and incorporate some additional information
ONE-STAGE DECISION MAKING
PROBLEM
´ In case of single stage decision problems calls for:

i. Identification of the course of action available to the decision

maker in the face of various possible events

ii. Developing a pay-off matrix

iii. Choosing a particular course of action

NUMERICAL
A Book store sells a particular book of tax laws for Rs.100. It
purchases the book for Rs.80 per copy. Since some of the tax laws
change every year, the copies unsold at the end of a year becomes
outdated and can be disposed of for Rs.30 each. According to past
experience, the annual demand for this book is between 18 and 23
copies.

à
 Order for this book can be placed only once in a year.

For this problem, since the annual demand varies between 18 and
23 copies, there are six possible events and course of actions.
Copies Possible
demanded strategies
E1:18 A1:18
E2:19 A2:19
E3:20 A3:20
E4:21 A4:21
E5:22 A5:22
E6:23 A6:23
Developing Pay-off And Regret Tables

´ A pay-off is a conditional value- a conditional profit, loss, or may be a

conditional cost.
´ A pay- off table thus represents the matrix of the conditional values
associated with all the possible combinations of the acts and events.
´ To construct the pay- off table:
 Let D = demand (events)
 Q = quantity (actions)
 P = profit
 P= (total revenue from the sale of all copies ± total cost of
procuring them)
´ 
ëë 

´ hen D•Q :

´ P = 100Q ± 80 Q
= 20 Q
It shows bookseller getting profit of Rs. 20 on the sale of each
quantity
´ 
ë 
´ 

´ P = 100 D + 30 (Q- D) ± 80 Q
P = 100D + 30 Q- 30 D- 80 Q
P = 70 D- 50 Q

HERE,
P= total revenue obtained from selling D copies + revenue
from the unsold copies ± total cost of buying Q copies.
Pay-off Table

‘ ‘
à 
à  à  A4:21 A4:22 A4:23
‘
E1: 18 360 310 260 210 160 110
E2: 19 360 380 330 280 230 180
E3: 20 360 380 400 350 300 250
E4: 21 360 380 400 420 370 320
E5: 22 360 380 400 420 440 390
E6: 23 360 380 400 420 440 460
Opportunity Loss or Regret Table
´ The resultant outcomes of various combinations of the acts and
events can also be expressed in terms of the OPPORTUNITY
LOSS.

´ | |

|is defined as the amount of pay-off

foregone by not adopting the optimal course of action-that would
give the highest pay-off, for each possible event.
Regret Table

‘ ‘
à 
à  à  à à à
‘
E1: 18 0 50 100 150 200 250
E2: 19 20 0 50 100 150 200
E3: 20 40 20 0 50 100 150
E4: 21 60 40 20 0 50 100
E5: 22 80 60 40 20 0 50
E6: 23 100 80 60 40 20 0
Decision Rules
´ The Pay-off matrix or Regret matrix serves adequate
representation of business decision situations, but all types of
decisions cannot be taken through them.

´ Therefore In Practical situations, Decisions to be taken are

followed by two decision rules:
i. Decision Under Uncertainty

ii. Decision Under Risk

Decision Rules
´ After Developing Pay-off Table or Regret Table, the Decision
Maker takes decision using certain Decision Rules on the basis of
which decision may be taken.

  
 



w Decision Under Uncertainty

w Decision Under Risk

 Decision Under UnCertainty
´ The decision situations where there is no way in which the
decision-maker can assess the probabilities of the various states of
nature is known as 6ë6
 
´ Types of Principle used:
i. Laplace Principle

ii. Maximin Principle

iii. Maximax Principle

iv. Hurwicz Principle

v. Savage Principle
Laplace Principle
´ Under this all the events are treated equally probable .

´ The Expected (Mean) value of Pay-off for each strategy

(events) is determined .

´ The strategy with highest mean value is adopted.

‘ ‘
 à 
à  à  à à à
‘
E1: 18 360 310 260 210 160 110
E2: 19 360 380 330 280 230 180
E3: 20 360 380 400 350 300 250
E4: 21 360 380 400 420 370 320
E5: 22 360 380 400 420 440 390
E6: 23 360 380 400 420 440 460
™           

‘à
 
     


Expected Pay-off for A2 is the Maximum, Decision- maker would

choose to buy 19 copies of books
Maxmin or Minimax Principle
´ Adopted by Pessimistic¶ Decision Maker

´  ë

´ Choose 
 

´ Among the Minimum Pay-off, Choose the Maximum Value.

´ The pessimistic is a negative thinker and move forward by

keeping in mind the losses.
á
ááá


á

‘ ‘
à 
à  à  à à à
‘
E1: 18 360 310 260 210 160 110
E2: 19 360 380 330 280 230 180
E3: 20 360 380 400 350 300 250
E4: 21 360 380 400 420 370 320
E5: 22 360 380 400 420 440 390
E6: 23 360 380 400 420 440 460
Maximax or Minimin Principle
´ Adopted by Optimists¶ Decision Maker

 ë
´ Choose 
  
´ Among the Maximum Pay-off, Choose the Maximum
Value.
´ The Optimists always desires a chance for the
maximum pay-off in the decision matrix.
 á


‘ ‘
à 
à  à  à à à
‘
E1: 18 360 310 260 210 160 110
E2: 19 360 380 330 280 230 180
E3: 20 360 380 400 350 300 250
E4: 21 360 380 400 420 370 320
E5: 22 360 380 400 420 440 390
E6: 23 360 380 400 420 440 460

The highest profit is 460, therefore strategyA6 i.e. buying 23 copies

will be decided for this particular principle
Hurwicz Principle
´ The Decision under this principle lies between the Max-min
Principle and the Maxi-max Principle.

´ It provides a mechanism by which different levels of optimism &

pessimism may be shown.

´ For this, a value is defined, â on a scale ranging from   .

while â =  indicates Extreme Pessimism while ù = represents

Extreme Optimism.
Steps involved

i. Multiply Maximum Profit of each Actions by ù.

ii. Multiply Minimum Profit of each event by  ù.

iii. Add the products of each events.

iv. Choose the maximum value alternative or strategy.

Problem
Assume ù = 0.6,  ù = 0.4

ACT MAX MIN CRITERIAN VALUE = ù Ô


Ô ù Ô

A1 360 360 0.6*360+0.4*360=360
A2 380 310 0.6*380+0.4*310=352
A3 400 260 344
A4 420 210 336
A5 440 160 328
A6 460 110 320

Under this Principle, A1 Strategy will be chosen as it gives

the maximum criterion value
Savage Principle
´ This Principle is based on the concept of regret & calls for
selecting the actions that minimizes the maximum regret known as
Principle of áá
 ! 

´  ë

w Derive Opportunity Loss Matrix from Pay-off Matrix.

w Choose the Maximum regret value for each actions

w Out of these, determine the strategy which minimizes the

maximum regret.
 á
 ! "
#

‘ ‘
à 
à  à  à à à
‘

E1: 18 0 50 100 150 200 250

E2: 19 20 0 50 100 150 200
E3: 20 40 20 0 50 100 150
E4: 21 60 40 20 0 50 100
E5: 22 80 60 40 20 0 50
E6: 23 100 80 60 40 20 0

Under this Principle, A2 Strategy will be chosen as it gives

the minimum regret value out of the maximum
DECISION UNDER RISK
´ The decision situations wherein the decision ± maker chooses to
consider several possible outcomes & the probabilities of their
occurrence can be stated are called 6ë  ë$

´ Types of Principle used:

i. Maximum Likelihood Principle

ii. Expectation Principle

MAXIMUM LIKELIHOOD PRINCIPLE
´ ëë  Probabilities for various outcome ± 0.05, 0.10,
0.30, 0.40, 0.10 & 0.05 respectively
‘ ‘
Probabi à 
à  à  à à à
‘ lity Pi

E1: 18 0.05  310 260 210 160 110

E2: 19 0.10 360
 330 280 230 180
E3: 20 0.30 360 380  350 300 250

E4: 21 0.40 360 380 400  370 320

E5: 22 0.10 360 380 400 420  390
E6: 23 0.05 360 380 400 420 440 
‘ ‘
 à 
à  à  à à à
‘

Maximu 
    
m Pay-off

Probabilit 0.05 0.10 0.30 0.40 0.10 0.05

y Pi

(Pi)* 1 3 1 0 1 44 3
Maximu
m Pay
ff

According to this Principle, A4 Strategy will be chosen i.e.

buying 1 Books.
EXPECTATION PRINCIPLE

´ Under Situation of Risk, most of the decision are taken on the

basis of expectation principle.

´  ë

´ Multiply each pay-off values with the probability assigned to

it.

´ After multiplying all the pay off, add these for each actions.

´ EPj = ™ (pi)(aij)
EXPECTED PAY-OFF (CALCULATION):
´ A1: 0.05 360+0.10 360+0.30 360+0.40 360+0.10 360+0.05 360 =
360
´ A2: 0.05 310+0.10 380+0.30 380+0.40 380+0.10 380+0.05 380 =
376.5
´ A3: 0.05 260+0.10 330+0.30 400+0.40 400+0.10 400+0.05 400 =
386
´ A4: 0.05 210+0.10 280+0.30 350+0.40 420+0.10 420+0.05 420 =
374.5
´ A5: 0.05 160+0.10 230+0.30 300+0.40 370+0.10 440+0.05 440 =
375
´ A6: 0.05 360+0.10 180+0.30 250+0.40 320+0.10 390+0.05 460 =
288.5
According to this Principle, A3 strategy will be chosen as it gives
the highest expected pay ±off.
 EXPECTED REGRET (CALCULATION):

´ A1: 0.05 0+0.10 20+0.30 40+0.40 60+0.10 80+0.05 100 = 51

´ A2: 0.05 50+0.10 0+0.30 20+0.40 40+0.10 60+0.05 80 = 34.5

´ A3: 0.05 100+0.10 50+0.30 0+0.40 20+0.10 40+0.05 60 = 25

´ A4: 0.05 150+0.10 100+0.30 50+0.40 0+0.10 20+0.05 40 = 36.5

´ A5: 0.05 200+0.10 150+0.30 100+0.40 50+0.10 0+0.05 20 = 76

´ A6:0.05 250+0.10 200+0.30 150+0.40 100+0.10 50+0.05 0= 122.5

 
‘
 

‘ ‘
Probabi à 
à  à  à à à
‘ lity Pi

E1: 18 0.05 0 50 100 150 200 250

E2: 19 0.10 20 0 50 100 150 200
E3: 20 0.30 40 20 0 50 100 150
E4: 21 0.40 60 40 20 0 50 100
E5: 22 0.10 80 60 40 20 0 50
E6: 23 0.05 100 80 60 40 20 0
E ecte 51 34.5 25 36.5 6 122.5

e ret

Accor in to this Princi le, A3 strate y will be chosen as it

ives the lowest e ecte o ortunity loss or re ret.
Bayesian Decision Rule
´ Ên this approach the optimal strategy is chosen using the expected
value criterion while the expected payoffs are calculated by using
posterior probabilities.

´ The preliminary or the prior information of the decision maker is

revised on the basis of some additional information

´ The prior probabilities are converted into the posterior

probabilities
Numerical
´ Suppose Delhi developers limited is a construction company has recently
acquired a piece of land in the city on which it plans to construct a
shopping complex the company has now to decide the size of the complex,
its considering three options:

 ! 


A1 Small 40
A2 Medium 60
A3 Large 100

Demand would be either "" or . The return from the project will obviously
depend on " !  "    and " "  
 
  
 
  #. The payoffs in thousands(rupees)
expected under the event action combinations, together with the estimated
probabilities of the likely demand are given here

  %
%#  6 
˜

˜ ˜

 ˜
˜
  

Vigh 0.4 1800 2200 4200

Low 0.6 1000 600 -1200

For a more informed decision, the company has a choice of

engaging a market research firm conducted a survey which
has the following outcomes as indicated as I1 and I2.

I1 : A Favorable report , indicating high demand for

condominiums
I2 : An unfavorable report,
report indicating a low demand for
condominiums

  6 &
Ê
˜˜ Ê
˜˜
E1 : Vigh 0.9 0.1
E2 : Low 0.2 0.8

The marketing research firm has asked for U $$ as the fee
for undertaking the study. How should the construction company
proceed??
PRIOR Analysis:
Calculation Of Expected Pay-off
Best strategy

  %
6 
%#  à

˜ à

 à
˜
  

E1: Vigh 0.4 1800 2200 4200
E2: Low 0.6 1000 600 -1200
Expected pay-off 1320 1240 960
 ´ Expected Pay-Off for A1 = 1800 0.4 + 1000 0.6 = 1320
´ Expected Pay-Off for A2 = 2200 0.4 + 600 0.6 = 1240
´ Expected Pay-Off for A3 = 4200 0.4 - 4200 0.6 = 960

´ Expected pay-off for A1, EP1 = 1320 thousand is maximum.

Thus according to the given probabilities of high and low
demand, the company should construct a small sized
complex.
EPPI and EVPI
´ In the previous table in the event of high demand for
condominiums, the best strategy is a large sized complex with
a pay-off of 4200 thousand and in case of low demand ,small
sixed complex involving a pay-off of 1000 thousand is the
best course.

´ So using these value we determine expected pay-off of

perfect information(EPPI).

  %
%#  
' 
',%
(U )***+
E1: Vigh 0.4 4200 1680
E2: Low 0.6 1000 600
EPPI = 2280

EVPI = EPPI ± EP = 2280 ± 1320 = ` 960 thousand

‘ = Expected Pay-Off...
‘  = Expected Pay-Off of Perfect Information...
‘  = Expected Value of Perfect Information...
Posterior Analysis
´ Revised decision of the company if it engages the marketing
research firm.

´ The approach involves revising the prior probabilities of high

and low demand to calculate posterior probabilities in the first
place, and then using thee posterior probabilities to reach a
decision. This is called posterior analysis.
Conditional Probabilities
w Given that if the state of nature eventually turns out to be a
high demand, what is the probability that the marketing
research study shall give a favorable report=0.90,
w This is the conditional probability, P(I1/E1)
It means;
P(I1/E1)=0.90
w Similarly,
 P(I1/E2)=0.20
 P(I2/E1)=0.10
 P(I2/E2)=0.80

.
Using the prior and conditional probability, we determine the total
probability that the research report will be favorable or unfavorable

  6 &
&-

%# & .

%#
E1 : Vigh 0.9 0.1
E2 : Low 0.2 0.8

P(I1) = P(E1 ŀ I1) + P(E2 ŀ I1)

± P(I1) = P(E1) P(I1/E1) + P(E2) P(I1/E2)
Substituting all values, we get
P(I1) = 0.4 0.9+0.6 0.20
= 0.36+0.12=0.48

Similarly,
P(I2) = P(E1 ŀ I2) + P(E2 ŀ I2)
± P(I2) = P(E1) P(I2/E1) + P(E2) P(I2/E2)
Substituting all values, we get
P(I1) = 0.4 0.1+0.6 0.80
= 0.04+0.48=0.52
Calculation Of Posterior Probabilities For The Events
E1 And E2 Under The Conditions

´ hen a favorable report is given (indicator I1)

‘    %    
‘ ## ## ## ##
ԑ Ô ‘ Ô Ë ‘ ԑ 

E1 0.4 0.90 0.36 0.36/0.48=0.7

5
E2 0.6 0.20 0.12 0.12/0.48=0.2
5

Using these posterior probabilities we will determine the optimal

course of action
 ‘ ## à  à à  
‘   
 

E1: High 0.75 1,800 2,200 4,200
Demand
E2: Low 0.25 1,000 600 -1,200
Demand
Expected Pay-Off 1,600 1,800 2,850

The Best Course of Action would be the Construction of a Large

Sized Shopping Complex given the report to be favorable. This has
expected pay-off of ` 2,850 thousandi
hen An Unfavorable Report Is Given ( Indicator I2):
Calculation Of Posterior Probabilities

‘    %    

‘ ## ## ## ##
ԑ Ô‘ ÔË ‘ ԑ
E1 0.4 0.10 0.04 0.04/0.52=0.0
769
E2 0.6 0.80 0.48 0.48/0.52=0.9
231
Total 0.52

Using posterior probabilities determining the optimal course of

action
 Event Probability A1: Small A2: A3: Large
Ei Pi Complex Medium Complex
Complex
E1: High 0.0769 1,800 2,200 4,200
Demand

E2: Low 0.9231 1,000 600 -1,200

Demand

Expected Pay-Off 1,061.52 723.04 -784.74

The expected pay off of the act A1 is largest : small sized shopping
complex
´ Favorable market research report is obtained

w Optimal course of action => A3

w Expected pay-off = ` 2,850 thousand

´ Unfavorable market research report is obtained

w Optimal course of action => A1

w Expected pay-off = ` 1,061.52 thousand

´ But these decisions are conditional in the sense that either of
them can be taken only when the nature of the report is
known.

´ Since there is likelihood of 0.48 for the report for the report to
be favorable, and 0.52 for it to be otherwise, we can get the
expected pay off value as shown below:
  ##    ‘ 

" à
I1: Favorable 0.48 2,850 1,368
Report
I2: Unfavorable 0.52 1,061.52 552
Report
Expected Pay Off 1,920

´ The Expected Pay-off of the optimal decision is ` 1,920

thousand, if it is based in the market research information, as
against ` 1,320 thousand without such information.
EVSI
´ Information on the basis of which prior probabilities are

revised is called sample information.

´ In view of the fact that the company would have to incur extra

cost by payment to the marketing research firm, it would like

to know the value to be placed on the information provided by

the research firm. So we determine the expected value of

sample information(EVSI)
´ EVSI = Expected Payoff with Sample Information ±

Expected Payoff without Sample information

´ Expected payoff with sample information = 1,920 thousand

´ Expected pay-off when no sample information = 1,320 thousand

EVSI = 1,920 - 1,320 = 600 thousand

´ The company, can therefore, pay an amount upto ` 600,000 to

the marketing research firm for the research report. Since it is

given here that the marketing research firm has asked a fee of

` 300,000, it is worth engaging it.

Efficiency of EVSI
´ EVPI for the given problem is 960 thousand.

´ However as the report is not always likely to yield correct

information, we can measure its efficiency by relating EVSI to
EVPI

´ Thus an efficiency index, EI, of sample information can be

obtained as follows:

EI=(EVSI/EVPI) 100
´ In our example,

EI=(600/960) 100= 62.5%

Thus the sample information is about 63% as efficient as perfect

information.

Calculation of the efficiency index enables the decision maker to

decide whether or not to seek the information and compare various
sources through which the needed information may be obtained
Multi-stage Decision Making :Decision Tree
´ Decision Tree is a graphic representation of the sequences of action-
event combinations available to the decision maker.

´ Multiple Decisions also called Sequential Decision, that are

characterized by a sequence of decisions.

´ A chance event occurs which in turn influences the next generation.

´ Evaluates the decision proceeding in a backward manner by evaluating

the best course of action at the later stages to decide the best action at
the earlier stages.

´ For this purpose we construct Decision Tree / Decision Flow Diagram.

´ Expectation Principle : Maximizes Profit or Minimizes Cost

 Problems???
´ An oil company has recently acquired rights in a certain area to
conduct surveys and tests drilling to lead to lifting oil .The
company has the choice to conduct further geological tests or to
carry out a drilling programme .on the known condition ,the
company estimates that there is a 70:30 chance of further tests
showing a µsuccess¶.
hether the test show the possibility of ultimate success or not or
even if no tests are undertaken at all, the company could still
pursue its drilling programme or alternatively consider selling its
rights to drill in the area.
If it carries out the drilling programme, the likelihood of final success
or failure is considered dependent on following stages:

´ If µsuccessful¶ tests have been carried out, the expectation of

success in drilling is 80:20.

´ If the tests include µfailure', then the expectation of success in

drilling is 20:80.

´ If no test have been carried out at all the expectation of success in

drilling is 55:45

Cost and revenues have been estimated for all possible outcomes and
the net present value of each is as follows:
 

  



  with prior tests. 100

without prior tests. 120
&
 with prior tests -50
without prior tests -40
 " 
prior tests show µsuccess¶ 65
prior tests show µfailure¶ 15
without prior tests 45
a)Draw up a decision tree diagram to represent the above
information.
b)Evaluate the tree in order to advice the management of the
company on its best course of action.
 65 -50
-40

2 0.8 100
120 success

test
3

drill success
100
1 0.2
45


15
-50
cecision node

Chance node
Evaluation of Decision Node 1:

   
 #  ‘


 

1. Drill Success 0.2 100 20

Failure 0.8 (50) (40)

Total (20)
    

On the basis of the criterion of maximization of the expected

profit, our decision would be to sell the site.
Evaluation of decision Node 2:
 à  
 #  ‘ 



1. sell 1.0 65 65

2. Drill Success 0.8 100 80

Failure 0.2 (50) (10)



Here the best course would be to go in for oil drilling.

 Evaluation of Decision Node 3:
   
 #  ‘ 



1. Drill Success 0.55 120 66

Failure 0.45 (40) (18)



2. Test Positive 0.7 70 49

Negative 0.3 15 4.5

 
3. Sell 1.0 45 45
 - 65 -50
-40

70 100
48 120 70
0.55

53.5 53.5

0.2 100
15 20
45
15
-50
 e may observe that the expected value of the alternative of

carrying out a test is Rs. 53.5 millions ,which is the highest of the

three .Therefore it is better to test before drilling.