Decision Theory

• In Many situations , we have to take important decisions

which involve a lot of reasoning and thinking . For example,
decision whether(i) to launch or not to launch a new
product in the market , (ii) to increase or not to increase
production of a manufacturing company etc are important
decisions. If there is only one course of action , no decision
is necessary and we choose the given course of action to
solve the problem.
• There are three elements in decision making.
• (i) Act (ii) states of nature (iii) Decision.
• Act: An act or strategy or decision alternative is the course
of action taken by the decision maker.
• Decision: A decision problem has at least two possible acts
(or decision alternative available)
• State of Nature: The outcome of any action depend on
certain factors which are often uncontrollable. These factors
are states of nature (or event state).
• Example:
A cement company can either produce 100 or 120 tons of cement per
month. The total production of this competitors can be either 10000 or
15,000 tones. If they produce 10,000 tones per month , the profit per tone is
Rs 120, but if they produce 15,000 tones, his profit per ton is only Rs 80.
Construct a payoff table.
Solutions: Here the acts are : a1 : produce 100 tons, (ii) a2: produce 120
tons , and the states of nature are:
(i) S1 : Production of competitor is 10,000 tons , and (ii) S2: production of
competitor is 15,000 tons.
TIME PAY OFF TABLE FOR THE GIVEN INFORMATIONS
State of nature
Acts S1: 10,000 tons S2: 15000 tons
a1: Produce 100 tons 100xRs 120= Rs 12,000 100xRs80= Rs 8,000
a2: Produce 120 tons 120x Rs 120=Rs 14,400 120xRs 80= Rs 9,600
• Decision Making Under Certainty
It is the process of selecting a decision alternative when the
states of nature are known. In this case , there will be only
one state nature column in the payoff and the action(act)
that yields the best payoff in the column if chosen.

• Decision Making under Uncertainty

It is the process of selecting a decision alternative (or act)
when the states of nature are not known. In this case , the
decision –maker first selects a criterion and then determines
which decision alternatives(ie act) is the best under the
chosen criterion.
 Learning Objectives
• Decision Making and its process.
• Decision Making Under Uncertainty (without
Probability)
• Decision making under risk ( With Probability)
• The success or failure that an individual
/organization experiences , depends to a large
extends on the ability of making appropriate
decision.
Making decision requires…
• An enumeration of feasible and viable
alternatives (courses of action/ strategies)
• The Projection of consequences associated with
different alternatives
• And measure of effectiveness (or an
objectives)by which most preferred alternatives
is identified.
• Decision theory provides an analytical and
systematic approach to the study of decision
making.
• In other words, decision theory provides a
method of natural decision making wherein
data concerning the occurrence of different
outcome (consequences) may be evaluated to
enable the decision maker to identify suitable
alternatives ( or course of action)
• Decision models useful in helping Decision-
makers make the best possible decisions are
classified according to the degree of
uncertainty.
• The scale of certainty can range from complete
certainty to complete uncertainty.
• The regions which falls between these extreme
points corresponds to decision making under
risk ( Probabilistic Problem)
Essential Characteristics of Decision Model

• Decision Alternatives ( Course of Action/

Strategies)- There is finite number of
decision alternatives available with decision
maker at each point in time when decision is
to be made.
• State of nature-
The possible future condition( consequences
/event) resulting from the choice of decision
alternative depends on certain factors beyond
the control of
• Pay-off
• A numerical value resulting from each possible
combination of alternatives and state of
nature is called as pay-off.
• The pay off values are always conditional
values because of unknown state of nature
• General form of pay off matrix
Steps in decision making approach
• Listing all possible state of nature which
occurs in the context of decision problem.
• Identification of all courses of action which are
available to the decision maker.
• Expressing the payoff (pij)resulting from each
pair of courses of action and state of nature.
• Choosing an appropriate course of action
from the given list on the basis of some
criterion that results in optimal pay off.
Decision making under risk
(With Probabilities
• Decision making under risk is probabilistic
decision making situation , in which more than
one state of nature exists and DM has
sufficient information to assign probability
values to their likely occurrences of each of
these states.
• Knowing the probability distribution of all the
states of nature, the best decision is to select
that course of action which has largest
expected payoff value.
 Decision making under risk
(With Probabilities)
1) The Expected Monetary Value
(EMV)Criterion
2) The Expected Value with Perfect
Information (EVPI)
3) Value of perfect Information (VPI)
4) The Expected Opportunity Loss
(EOL)Criterion
 The Expected Monetary Value
(EMV)Criterion
• It is possible to obtain probability estimates
for each state of nature in decision-making
situations. We use the expected value
criterion (used in Statistics) to identify the
best decision alternative.
 • The expected monitory value EMV is calculated by multiplying
each decision outcome (payoff value) for each state of nature
by the probability of its occurrence.
Course of Action (Strategies)
Probability Large- Medium- Small - No-
State of Nature Sized Sized Sized Facility(S4)
Facility Facility Facility(S
(S1) (S2) 3)
Strong Market 0.30 550000 300000 200000 0
(Rs Profit) N1
Fair Market (Rs 040 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. 0.30 -310000 -100000 -32000 0
Profit N3
Expected Monetary 116000 111600 90400 0
Value
S1=550000*0.30+110000*0.40+(-310000)*0.30
=165000+44000-93000
=116000.
S2=300000*0.30+129000*040+(-100000)*0.30
=90000+51600-30000
=111600
S3=200000*0.30+100000*0.40+(-32000)*0.30
=60000+40000-9600
=90400
 The Expected Value of Perfect
Information
Course of Action (Strategies)
Probability Large- Medium- Small - No-
State of Nature Sized Sized Sized Facility(S4
Facility Facility Facility(S )
(S1) (S2) 3)
Strong Market 0.30 550000 300000 200000 0
(Rs Profit) N1
Fair Market (Rs 040 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. 0.30 -310000 -100000 -32000 0
Profit N3
Expected 116000 111600 90400 0
Monetary Value
• If we examine the best two EV's from the
above example,
• EMV(S1)=116000
• EMV(S2)=111600
There is difference of Rs. 4400.
But if we know there is a poor market
condition, a choice of S2 would be a better
choice than S1.
• Then, how can we figure out if such a condition
may occur? One way is to hire a consultant to
do some research on future market conditions.
• But it would be foolish for us to pay a
consultant fee more than what we could gain
in extra profit from the consultant's
information.
• The information has some maximum value that
represents the limit of what we would be
willing to spend. This value of information can
be computed as an expected value, referred to
as the expected value of perfect information
(EVPI).
• "Perfect information" leads a decision maker
to no regret (the right decision) under each
state of nature.
 The Expected Value of Perfect
Information
Course of Action (Strategies)
Probability Large- Medium- Small - No-
State of Nature Sized Sized Sized Facility(S4
Facility Facility Facility(S )
(S1) (S2) 3)

Strong Market 0.30 550000 300000 200000 0

(Rs Profit) N1

Fair Market (Rs 040 110000 129000 100000 0

Profit) N2

Poor Market ( Rs. 0.30 -310000 -100000 -32000 0

Profit N3

EMV 116000 111600 90400 0

• Using perfect information, the expected value
of the decision is
=550000*0.30+129000*0.40+0*0.30
=216600
• The expected value of perfect information is
the maximum amount that would be paid to
gain information that would result in a
decision better that the one made without
perfect information.
• The expected value decision without perfect
information is to construct a large-sized facility
• EMV(S1)=116000
EVPI =the expected value given perfect
information -the expected value without
perfect information.
EVPI=216000-116000
• 100,600 is the maximum amount that we
would pay to purchase perfect information
from other sources like a consultant and
marketing surveys. It is rare for us to obtain
perfect information in general.
• It is noted that the expected value of perfect
information is the same as the expected
opportunity loss for the decision alternative
selected.
• EVPI = best EOL.
• This is always the case because if we use
perfect information, there is no regret.
 The Expected Opportunity Loss
(EOL)Criterion
• An alternative approach to maximizing EMV is
to minimize Expected opportunity loss.
• The EOL is defined as the difference between
the highest payoff for the state of nature and
the actual profit obtained for the particular
course of action taken.
• In other words, EOL is the amount of payoff
that is lost by not selecting the course of
action that has greatest payoff for the state of
nature that actually occures.
• The course of action due to which EOL is
minimum is recommended.
 Course of Action (Strategies)

Probability Large-Sized Medium-Sized Small -Sized No-Facility(S4)

State of Nature Facility (S1) Facility Facility(S3)
(S2)

Strong 0.30 550000 300000 200000 0

Market (Rs
Profit) N1
Fair Market 0.40 110000 129000 100000 0
(Rs Profit)
N2

Poor Market 0.30 -310000 -100000 -32000 0

( Rs. Profit)
N3
 Course of Action (Strategies)

Probability Large- Medium- Small - No-Facility(S4)

State of Sized Sized Sized
Nature Facility Facility Facility(S3
(S1) (S2) )
Strong 0.30 (550000- (550000- (550000- (550000-0)
Market (Rs 550000) 3000000) 200000) 550000
Profit) N1 =0 =250000 =350000
Fair 0.40 (129000- (129000- (129000- (129000-0)
Market (Rs 110000) 129000) 100000) =129000
Profit) N2 =19000 =0 =29000

Poor 0.50 0-(- 0-(-100000) 0-(-32000) (0-0)

Market ( 310000) =100000 =32000 =0
Rs. Profit) =310000
N3
EOL 100600 105000 126200
EOL (S1)=0*0.30+19000*0.40+310000*0.30
=100600

EOL(S2)=250000*0.30+0*0.40+100000*0.30
=105000

EOL(S3)=3500000*0.30+29000*0.40+35000*0.30
=126200

EOL(S4)=550000*0.30+129000*0.40+0*0.30
=216600
The best decision results from minimizing the regret.
In this case, the decision is a "large-sized facility.“
S1 strategy is recommended
Q. A florist , in order to satisfy the needs of a
number of regular and sophisticated customers,
stocks highly perishable flowers. A dozen flowers
cost Rs.30 and sells at Rs.100. Any flower not
sold on the day is worthless.
• The Demand Distribution in dozens of flowers is
as follows.

Demand 1 2 3 4

Probability 0.2 0.3 0.3 0.2

• How many flowers he should stock in order to

maximize the expected net profit?
 SP=100 ;CP=30; Profit=70
Demand(in Probabili Stocks(in Dozens)
Dozens) ty

1(S1) 2(S2) 3(S3) 4(S4)

1 (N1) 0.2 1*70=70 70*1- 70*1- 70*1-3*30=

30*1=40 30*2=10 -20

2 (N2) 0.3 1*70=70 70*2=140 70*2- 70*2-

30*1=110 30*2=80

3 (N3) 0.3 1*70=70 70*2=140 70*3=210 70*3-

30*1=180

4(N4) 0.2 1*70=70 70*2=140 70*3=210 70*4=280

Demand(in Probability Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 0.2 70 40 10 -20

2 (N2) 0.3 70 140 110 80

3 (N3) 0.3 70 140 210 180

4(N4) 0.2 70 140 210 280

EMV (profit) 70 120 140 130

• Florist should stock 3 dozen flowers each day to get

maximum profit of 140.
S1=70*0.2+70*0.3+70*.03+70*0.2=70
S2=40*0.2+140*0.3+140*0.3+140*0.2=120
S3=10*0.2+110*0.3+210*0.3+210*0.2=140
S4=(-20)*0.3+80*0.3+180*0.3+280*0.2=130
Find EMV, EVPI, VPI and EOL.
 Decision Making Under
Uncertainty and without probabilities
The Maximax Criterion
The Maximin Criterion
The Laplace Criterion (equally likely )
The Hurwitz Criterion
The Minimax Regret criterion
 i
Course of Action (Strategies)
Large-Sized Medium- Small - No-
Facility (S1) Sized Sized Facility(S4)
Facility Facility(S
State of Nature (S2) 3)
Strong Market (Rs 550000 300000 200000 0
Profit) N1
Fair Market (Rs 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. -310000 -100000 -32000 0
Profit N3
 The Maximax Criterion
(Optimistic Decision)
• DM ensures that (S)he should not miss the
opportunity to achieve the largest possible
profit( Maximax ) and the lowest possible
costs.(Minimin)

• The maximax criterion assumes of an

optimistic decision maker.
 Course of Action (Strategies)
Large-Sized Medium- Small - No-
Facility (S1) Sized Sized Facility(S4)
Facility Facility(S3)
State of Nature (S2)
Strong Market (Rs 550000 300000 200000 0
Profit) N1
Fair Market (Rs 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. -310000 -100000 -32000 0
Profit N3
Max 500000 300000 200000 0
Maximax
 The Maximin Criterion
(Pessimistic Decision)
• DM is conservative about future and always
anticipates worst possible outcome ( Min-
Profit / Max –costs)

• The maximin criterion assumes of an

Pessimistic decision maker.
 Course of Action (Strategies)
Large-Sized Medium- Small - No-
Facility Sized Facility Sized Facility(S4)
(S1) (S2) Facility(S
State of Nature 3)
Strong Market (Rs 550000 300000 200000 0
Profit) N1
Fair Market (Rs 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. -310000 -100000 -32000 0
Profit N3
-310000 -100000 -32000 0
Min
Maximin
 Laplace/ Equally likely criterion
• Since the probabilities of state of nature are
not known, it is assumed that all state of
nature occurs with equal probability.
• The Laplace criterion is also called the equal
likelihood criterion, which equally weights the
possibility of occurrence of each state of
nature.
 Course of Action (Strategies)

Large-Sized Medium-Sized Small - No-

State of Nature Facility (S1) Facility Sized Facility(S4)
(S2) Facility(S3)
Strong Market (Rs 550000 300000 200000 0
Profit) N1
Fair Market (Rs 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. -310000 -100000 -32000 0
Profit N3
=(55000+1100 =(30000+129000 =(200000+10 =(0+0+0)/3
000- 0-100000)/3 0000-
310000)/3 32000)/3
Expected Payoff =116666.67 =109666.67 89333.33 0
 The Hurwitz Criterion
• The rational decision maker should be neither optimistic
nor pessimistic and therefore must display a mixture of
both.
• The coefficient of optimism (α) is used to measure of the
DM’s degree of optimism.
• The α lies between 0 to 1
Step 1 – decide the coefficient of optimism (α)
Step 2 – Find the max (M) and min (m) payoff corresponding
to courses of action/strategy.
Step 3. Calculate the Expected value of payoff for each
strategy as: Expected Value = M α + (1- α ) m
Step 4: Identify the maximum pay-off among all these
expected values.
Step 5: The corresponding strategy to this maximum pay-off ,
is the optimal strategy.
 Coefficient of Optimism α=0.60
Course of Action (Strategies)

Large-Sized Medium-Sized Small -Sized No-

State of Nature Facility (S1) Facility Facility(S3) Facility(S4)
(S2)
Strong Market (Rs 550000 300000 200000 0
Profit) N1
Fair Market (Rs 110000 129000 100000 0
Profit) N2
Poor Market ( Rs. -310000 -100000 -32000 0
Profit N3
Max (M) 550000 300000 200000 0

Min (m) -310000 -100000 -32000 0

Expected Payoff =550000*0.60+( =300000*0.60+(- =200000*0.60 =0*0.60+0*0
-310000)*0.40 100000)*.40 +(- .40
32000)*0.40
Expected Payoff 206000 140000 107200 0
 The Minimax Regret criterion

• The criterion is known as opportunity loss

decision criterion because DM feel regret after
adopting a wrong course of action resulting
into opportunity loss of payoff.
• The working method Is as follows
• A)from the given payoff matrix,, develop an
opportunity loss matrix as follows,
• i)find the best payoff corresponding to each
state of nature.
• (ii)substract all other entries (payoff values) in
that row from this table.
• For each course of action ,identify the max
regret.
• C) select the course of action with smallest
anticipated oportunity loss value.
 Course of Action (Strategies)
Large-Sized Medium-Sized Small -Sized No-Facility(S4)
State of Nature Facility (S1) Facility Facility(S3)
(S2)

Strong Market (Rs 550000 300000 200000 0

Profit) N1

Fair Market (Rs 110000 129000 100000 0

Profit) N2

Poor Market ( Rs. -310000 -100000 -32000 0

Profit) N3
 Course of Action (Strategies)

Large- Medium-Sized Small -Sized No-Facility(S4)

State of Nature Sized Facility Facility(S3)
Facility (S2)
(S1)
Strong Market (550000- (550000- (550000- (550000-0)
(Rs Profit) N1 550000) 3000000) 200000) 550000
=0 =250000 =350000
Fair Market (129000- (129000- (129000- (129000-0)
(Rs Profit) N2 110000) 129000) 100000) =129000
=19000 =0 =29000

Poor Market ( 0-(- 0-(-100000) 0-(-32000) (0-0)

Rs. Profit) N3 310000) =100000 =32000 =0
=310000
Maximum 310000 250000 350000 550000
Regret
Minimax
• Each of the highlighted payoffs is the
maximum for each alternative. The minimum
among the highlighted maximum is 250,000
for a "medium-sized facility." The decision
should be to construct a medium-sized facility.
 Summary of Criteria Results
Criteria Decision

The Maximax Criterion Large –Sized Facility

The Maximin Criterion No-Faciilty

The Laplace Criterion (equally Large-Size Facility

likely )
The Hurwitz Criterion Small-sized facility

The Minimax Regret criterion Medium-Size Facility

Exercise:
Consider the following Profit Table:

Strate States
gies
N1 N2 N3 N4

S1 30 10 10 8

S2 40 -15 5 7

S3 50 20 -6 10
 Exercise
• Q. A florist , in order to satisfy the needs of a
number of regular and sophisticated
customers, stocks highly perishable flowers.
A dozen flowers cost Rs.30 and sells at
Rs.100. Any flower not sold on the day is
worthless.
Demands for the flower N1-1 dozens ; N2: 2
Dozens; N3: 3 Dozens;N4: 4 dozens.
How many dozens of flowers he should stock
to maximize his expected profit?
 SP=100 ;CP=30; Profit=70
Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2) 3(S3) 4(S4)

1 (N1) 1*70=70 70*1-30*1=40 70*1-30*2=10 70*1-3*30= -20

2 (N2) 1*70=70 70*2=140 70*2-30*1=110 70*2-30*2=80

3 (N3) 1*70=70 70*2=140 70*3=210 70*3-30*1=180

4(N4) 1*70=70 70*2=140 70*3=210 70*4=280

Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 70 40 10 -20

2 (N2) 70 140 110 80

3 (N3) 70 140 210 180

4(N4) 70 140 210 280

Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 70 40 10 -20

2 (N2) 70 140 110 80

3 (N3) 70 140 210 180

4(N4) 70 140 210 280

MaxiMax 70 140 210 280

MaxiMin 70 40 10 -20
Demand Stocks(in Dozens)
(in Dozens)
1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 70 40 10 -20

2 (N2) 70 140 110 80

3 (N3) 70 140 210 180

4(N4) 70 140 210 280

Expected =(70+70+70+ =(40+140+140+14 =(10+110+210+2 =(-

profit 70)/4 0)/4 10)/4 20+80+180+280)/4
(Laplace) =70 =115 =135 =130
 α= 0.80
Demand(in Stocks(in Dozens)
Dozens)
1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 70 40 10 -20

2 (N2) 70 140 110 80

3 (N3) 70 140 210 180

4(N4) 70 140 210 280

Max(M) 70 140 210 280

Min (m) 70 40 10 -20

=Mα+m(1-α) =70*0.80+70*0. =140*0.80+40* =210*0.80+10* =280*0.80+(-

20 0.20 0.20 20)*0.20
=56+14 =112+8 =168+2 =220
=70 =120 =170
 The Minimax Regret criterion
Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 70 40 10 -20

2 (N2) 70 140 110 80

3 (N3) 70 140 210 180

4(N4) 70 140 210 280

Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) =70-70 =70-40 =70-10 =70-(-20)

=0 =30 =90

2 (N2) =140-70 =170-140 =170-110 =170-80

=70 =30 =60 =90

3 (N3) =210-70 =210-140 =210-210 =210-180

=140 =70 =0 =30

4(N4) =280-70 =280-140 =280-210 =280-280+=0

=210 =140 =70
Demand(in Stocks(in Dozens)
Dozens)

1(S1) 2(S2)15 3(S3) 4(S4)

1 (N1) 0 30 60 90

2 (N2) 70 30 60 90

3 (N3) 140 70 0 30

4(N4) 210 140 70 0

Maximum 210 140 70 90

Regret Minimax
 Summary
• decision theory provides a method of natural
decision making wherein data concerning
the occurrence of different outcome
(consequences) may be evaluated to enable
the decision maker to identify suitable
alternatives ( or course of action).
• Various criterion of DM under Uncertainty
and Risk enables decision makers to
undertake specific course of action.
 Question
• What is decision making? Explain steps in
decision making.
• Discuss various techniques/criterion of
decision making under uncertainty.
• Discuss various criterion of decision making
under Risk.
• Numericals on Decision making under
uncertainty and Risk.