Decision System and Analysis

Decision Analysis

A.A.B. Dinariyana
Learning
objectives
01 List the steps of the decision-making process

Describe the types of decision-making

02 environment

Make decision under uncertainty where the

03 probability is not known

Use probability values to make decision under

04 risk

Reading: Render, B., Stair, R.M. & Hanna (2012). Quantitative Analysis for
Management. (11th Ed). Pearson Education. Harlow, England
2
Course outline
The six steps in decision
02
01 Introduction theory

Types of decision making

Decision making under 03
uncertainty 04 environments

06 Decision Tree
05 Decision under risk

Sensitivity Analysis 07
3
Introduction
• Decision theory is an analytic and systematic approach to the study of decision making
• A good decision is one that is based on logic, considers all available data and possible
alternatives, and the quantitative approach described
Few
Under Alternatives
Certainty

Under # of
Environment
Uncertainty Alternatives

Large
Under Risk Number of
Alternatives

4
The six steps in decision-making
Step 1 Clearly define the problem at hand

Step 2 List the possible alternatives

Step 3 Identify the possible outcomes or states of nature

List the payoff or profit of each combination of alternatives and

Step 4 outcomes

Step 5 Select one of the mathematical decision theory models

Step 6 Apply the model and make your decision

5
The six steps in decision-making
Clearly define the problem at hand
STATE OF NATURE
Ex: Expand by manufacturing and
marketing a new product ALTERNATIVE
FAVORABLE MARKET UNFAVORABLE
List the possible alternatives ($) MARKET ($)

Construct a large plant 200,000 –180,000

Identify the possible outcomes or
states of nature
Construct a small plant 100,000 –20,000
List the payoff or profit of each
combination of alternatives and Do nothing 0 0
outcomes

Select one of the mathematical All Conditional value

decision theory models Will be
Apply the model and make your discussed in
decision the next
section
6
Types of decision-making environments
Decision Maker knows with certainty the
Decision Making under Certainty consequences of every alternative of decision
choice

Decision Decision Maker does not know the probabilities of

Decision Making Under Uncertainty the various outcomes
Making

Decision Making Under Risk Decision Maker does know the probabilities of the
various outcomes

7
Decision-making under uncertainty

Maximax (Optimistic)

Decision-making under
Maximin (Pessimistic)
uncertainty: When decision
maker cannot assess the
Criterion of Realism
outcome probability with (Hurwics)
confidence or when virtually
no probability data are Equally likely (Laplace)
available
Minimax Regret

8
Maximax (optimistic)
Used to find the alternative that maximizes the maximum payoff
• Locate the maximum payoff for each alternative
• Select the alternative with the maximum number

State Of Nature
Alternative Maximum In A Row ($)
Favorable Market ($) Unfavorable Market ($)

Construct a large plant 200,000 –180,000 200,000

Best choice
Construct a small plant 100,000 –20,000 100,000

Do nothing 0 0 0

9
Maximin (pessimistic)
Used to find the alternative that maximizes the minimum payoff
• Locate the minimum payoff for each alternative
• Select the alternative with the maximum number

State Of Nature
Alternative Minimum in A Row ($)
Favorable Market ($) Unfavorable Market ($)

Construct a large plant 200,000 –180,000 –180,000

Construct a small plant 100,000 –20,000 –20,000

Do nothing 0 0 0

Best choice
10
Criterion of realism (Hurwicz)
A weighted average compromise between optimistic and pessimistic
o Select a coefficient of realism a
o Coefficient is between 0 and 1
o A value of 1 is 100% optimistic
o Compute the weighted averages for each alternative
o Select the alternative with the highest value

Weighted average =a (maximum in row)

+ (1 – a)(minimum in row)

11
Criterion of realism (Hurwicz)
• For the large plant alternative using a = 0.8
(0.8)(200,000) + (1 – 0.8)(–180,000) = 124,000
• For the small plant alternative using a = 0.8
(0.8)(100,000) + (1 – 0.8)(–20,000) = 76,000
State Of Nature
CRITERION OF REALISM
Alternative
Favorable Market ($) Unfavorable Market ($) (a = 0.8)$

Construct a large plant 200,000 –180,000 124,000

Best choice
Construct a small plant 100,000 –20,000 76,000

Do nothing 0 0 0

12
Equally likely (Laplace)
Considers all the payoffs for each alternative
• Find the average payoff for each alternative
• Select the alternative with the highest average

State Of Nature
Alternative Row Average ($)
Favorable Market ($) Unfavorable Market ($)

Construct a large plant 200,000 –180,000 10,000

Construct a small plant 100,000 –20,000 40,000

Best choice
Do nothing 0 0 0

13
Minimax regret
Based on opportunity loss or regret, the difference between the optimal
profit and actual payoff for a decision
• Create an opportunity loss table by determining the opportunity loss for not choosing
the best alternative
• Opportunity loss is calculated by subtracting each payoff in the column from the best
payoff in the column
• Find the maximum opportunity loss for each alternative and pick the alternative with
the minimum number

14
Minimax regret
Opportunity State Of Nature
Loss Tables Favorable Market ($) Unfavorable Market ($)
200,000 – 200,000 0 – (–180,000)
200,000 – 100,000 0 – (–20,000)
200,000 – 0 0–0

State Of Nature
Maximum in a Row
Alternative Favorable Market ($) Unfavorable Market ($)
Construct a large plant 0 180,000 180,000
Construct a small plant 100,000 20,000 100,000 Minimax
Do nothing 200,000 0 200,000

15
Decision-making under risk
Expected Monetary o Probability is known
Value (EMV) o Selection based on Maximum EMV

Decision-making
when there are
Expected Value of o Probability is known
several possible Perfect Information
o Selection based on upper bound
value on what to pay for
states of nature and (EVPI information
we know the
probabilities
o Probability is known
associated with Expected Opportunity
o Selection based on Minimum EOL
Loss (EOL)
each possible state

o To investigate how the decision

Sensitivity Analysis may change with different input

16
 Expected monetary value (EMV)
In the case of Thomson Lumber, the expected
monetary value for each alternative can be
calculated by using the following formula
n
EMV (alternative i) = å payoff S j • P(S j )
j=1
EMV (alternative i) = (payoff of 1st state of nature)
x (probability of 1st state of nature)
+ (payoff of 2nd state of nature)
x (probability of 2nd state of nature)
+ … + (payoff of nth state of nature)
x (probability of nth state of nature)

o Each market has a probability of 0.50 State Of Nature

o Which alternative would give the highest EMV? Alternative Favorable Unfavorable EMV ($)
Market ($) Market ($)
EMV (large plant) = (0.50)($200,000) + (0.50)(–$180,000) Construct a large
200,000 –180,000 10,000
plant
= $10,000
Construct a small
EMV (small plant) = (0.50)($100,000) + (0.50)(–$20,000) 100,000 –20,000 40,000
plant
= $40,000 Do nothing 0 0 0
Best choice
EMV (do nothing) = (0.50)($0) + (0.50)($0) Probabilities 0.50 0.50
= $0
17
Expected value of perfect information (EVPI)
How much money is Thompson
Lumber willing to pay for perfect
information?
The most Thompson Lumber
would be willing to pay is
determined by its EVPI
Expected Value of Perfect Information
• Perfect information can change decision
making under risk into decision making under
certainty
o EVPI places an upper bound on what you
should pay for additional information
o EVwPI is the long run average return if we
have perfect information before a decision is
made
EVPI = Expected Value with Perfect Information
– Expected Value without Perfect
Information
EVPI = EVwPI – Maximum EMV
18
 Expected value of perfect information (EVPI)
Best pay off for Best pay off for 2nd
1st state of nature state of nature
The expected value with perfect information is the
expected or average return, in the long run, if we have State Of Nature
perfect information before a decision has to be made. Alternative EMV ($)
Favorable Unfavorable
Market ($) Market ($)
EVwPI = (best payoff for 1st state of nature) Construct a large plant 200,000 –180,000 10,000
x (probability of 1st state of nature)
+ (best payoff for 2nd state of nature) Construct a small plant 100,000 –20,000 40,000
x (probability of 2nd state of nature)
+ … + (best payoff for nth state of nature) Do nothing 0 0 0
x (probability of nth state of nature) With Perfect
200,000 0 100,000
Information
Probabilities 0.50 0.50 EVwPI
o Scientific Marketing, Inc. offers analysis
that will provide certainty about market o Best alternative for favorable state of nature is build a large plant with a payoff of
conditions (favorable) $200,000 and Best alternative for unfavorable state of nature is to do nothing
with a payoff of $0
o Additional information will cost $65,000 EVwPI = ($200,000)(0.50) + ($0)(0.50) = $100,000

o Is it worth purchasing the information? o The maximum EMV without additional information is $40,000
EVPI = EVwPI – Maximum EMV So the maximum Thompson
should pay for the additional
= $100,000 - $40,000
information is $60,000
= $60,000 19
 Expected opportunity loss
State Of Nature
o Expected opportunity loss (EOL) is
Alternative Favorable Unfavorable EOL
the cost of not picking the best
solution Market ($) Market ($)
Construct a large plant 0 180,000 90,000
o First construct an opportunity loss
table Construct a small plant 100,000 20,000 60,000
Do nothing 200,000 0 100,000
o For each alternative, multiply the
opportunity loss by the probability Probabilities 0.50 0.50 Minimum EOL
of that loss for each possible
outcome and add these together
o Minimum EOL will always result in EOL (large plant) = (0.50)($0) + (0.50)($180,000)
the same decision as maximum = $90,000
EMV EOL (small plant) = (0.50)($100,000) + (0.50)($20,000)
o Minimum EOL will always equal = $60,000
EVPI EOL (do nothing) = (0.50)($200,000) + (0.50)($0)
= $100,000

20
 Sensitivity analysis
EMV Values

$300,000

o Sensitivity analysis examines how our decision $200,000

Point 2 EMV (large plant)
might change with different input data
$100,000 EMV (small plant)
o For the Thompson Lumber example Point 1

0 EMV (do nothing)

P = probability of a favorable market 0.167 0.615 1
–$100,000 Values of P
(1-P) = probability of an unfavorable market
–$200,000
EMV(Large Plant) = $200,000P – $180,000)(1 – P)
= $200,000P – $180,000 + $180,000P Point 1:
= $380,000P – $180,000 EMV(do nothing) = EMV(small plant)
20,000
EMV(Small Plant) = $100,000P – $20,000)(1 – P) 0 = $120,000 P - $20,000 P= = 0.167
120,000
= $100,000P – $20,000 + $20,000P
Point 2:
= $120,000P – $20,000 EMV(small plant) = EMV(large plant)
EMV(Do Nothing) = $0P + 0(1 – P) $120,000 P - $20,000 = $380,000 P - $180,000

= $0 160,000
P= = 0.615
260,000
21
Sensitivity analysis
Range of P
Best Alternative
Values
EMV Values Do nothing Less than 0.167
Construct a small plant 0.167 – 0.615
$300,000
Construct a large plant Greater than 0.615

$200,000 EMV (large plant)

Point 2

$100,000 EMV (small plant)

Point 1

0 EMV (do nothing)

0.167 0.615 1
–$100,000 Values of P

–$200,000
22
Decision trees
• Any problem that can be presented in a decision table can also be
graphically represented in a decision tree
• Decision trees are most beneficial when a sequence of decisions must be
made
• All decision trees contain decision points or nodes and state-of-nature
points or nodes
• A decision node from which one of several alternatives may be chosen
• A state-of-nature node out of which one state of nature will occur

23
Five steps to decision tree analysis
Define the problem

Structure or draw the decision tree

Tree Assign probabilities to the states of nature

Analysis
Estimate payoffs for each possible combination of
alternatives and states of nature

Solve the problem by computing expected monetary

values (EMVs) for each state of nature node

24
24
Structure of decision trees
o Trees start from left to right
o Represent decisions and outcomes in sequential order
o Squares represent decision nodes
o Circles represent states of nature nodes
o Lines or branches connect the decisions nodes and the states of nature

25
 Thompson’s decision tree

EMV for Node 1 = (0.5)($200,000) + (0.5)(–$180,000)

= $10,000
A State-of-Nature Node Payoffs
Favorable Market (0.5)
Favorable Market $200,000
Alternative with best
A Decision Node EMV is selected 1
1 Unfavorable Market (0.5)
Unfavorable Market –$180,000
ct t
ct t stru lan
tru lan n P
s
n P Co rge
Co rge La
La Favorable Market (0.5)
Favorable Market $100,000
Construct
Construct
2
2
Small Plant Unfavorable Market (0.5)
Small Plant Unfavorable Market –$20,000
Do Do
No
No
th th
ing EMV for Node 2 = (0.5)($100,000)
ing + (0.5)(–$20,000)
= $40,000

$0

26
 Thompson’s complex decision tree
1. Given favorable survey results,
EMV(node 2) = EMV(large plant | positive survey)
= (0.78)($190,000) + (0.22)(–$190,000) = $106,400
First Decision Second Decision Payoffs
Point Point EMV(node 3) = EMV(small plant | positive survey)
Favorable Market (0.78) = (0.78)($90,000) + (0.22)(–$30,000) = $63,600
$190,000
2 Unfavorable Market (0.22) EMV for no plant = –$10,000
ant –$190,000
e Pl
Larg Small Favorable Market (0.78)
$90,000 2. Given negative survey results,
5) Plant
3 Unfavorable Market (0.22)
0.4
–$30,000 EMV(node 4) = EMV(large plant | negative survey)
(
ey lts le No Plant = (0.27)($190,000) + (0.73)(–$190,000) = –$87,400
rv
Su Resu orab –$10,000
v EMV(node 5) = EMV(small plant | negative survey)
1 Surv Fa Favorable Market (0.27)
ey $190,000
R (0 4 Unfavorable Market (0.73)
= (0.27)($90,000) + (0.73)(–$30,000) = $2,400
Ne esu .55 ant –$190,000
e Pl
y

EMV for no plant = –$10,000

ve

ga lts )
La rg Favorable Market (0.27)
ur

tiv Small $90,000

tS

e 5 Unfavorable Market (0.73)

ke

Plant –$30,000
ar

3. Compute the expected value of the market survey,

M

No Plant
t
uc

–$10,000
nd

EMV(node 1) = EMV(conduct survey)

Co

= (0.45)($106,400) + (0.55)($2,400)
Do Favorable Market (0.50) = $47,880 + $1,320 = $49,200
Not $200,000
Con 6 Unfavorable Market (0.50)
duc ant –$180,000 4. If the market survey is not conducted,
t Su e Pl
rvey Larg Small Favorable Market (0.50)
$100,000
7 Unfavorable Market (0.50) EMV(node 6) = EMV(large plant)
Plant –$20,000
= (0.50)($200,000) + (0.50)(–$180,000) = $10,000
No Plant
$0 EMV(node 7) = EMV(small plant)
= (0.50)($100,000) + (0.50)(–$20,000) = $40,000
EMV for no plant = $0
5. Best choice is to seek marketing information

27
Thompson’s complex decision tree
First Decision Second Decision Payoffs
Point Point
$106,400 Favorable Market (0.78)
$190,000
t Unfavorable Market (0.22)
P la n –$190,000

$106,400
e
Larg $63,600 Favorable Market (0.78)
$90,000
Small
5) Unfavorable Market (0.22)
.4 Plant –$30,000
y (0
e lts le No Plant
rv b
Su esu ora –$10,000
R av
Su F –$87,400
rv Favorable Market (0.27)
ey $190,000
R (0 Unfavorable Market (0.73)
Ne esu .5 5
P la n
t
ey

–$190,000
ga lts ) e
Larg
rv

$2,400 Favorable Market (0.27)

$2,400
tiv
Su

e Small $90,000
t

Unfavorable Market (0.73)

ke

Plant –$30,000
ar
M
ct

No Plant
–$10,000
du
on
$49,200
C

Do $10,000 Favorable Market (0.50)

Not $200,000
Co n Unfavorable Market (0.50)
duc a nt
t Su e Pl –$180,000
$40,000

rvey Larg $40,000 Favorable Market (0.50)

$100,000
Small
Plant Unfavorable Market (0.50)
–$20,000
No Plant
$0

28
Expected value of sample information

Thompson wants to know the actual value of doing the survey

Expected value Expected value

with sample of best decision
EVSI = information, assuming
–
without sample
no cost to gather it information

= (EV with sample information + cost) – (EV without sample information)

EVSI = ($49,200 + $10,000) – $40,000 = $19,200

29
 Sensitivity analysis
How sensitive are the p = probability of a favorable survey result
decisions to changes in (1 – p) = probability of a negative survey result
the probabilities?
EMV(node 1) = ($106,400)p +($2,400)(1 – p)
o How sensitive is our decision
to the probability of a = $104,000p + $2,400
favorable survey result?
We are indifferent when the EMV of node 1 is the
o That is, if the probability of a
same as the EMV of not conducting the survey,
favorable result (p = .45) $40,000
where to change, would we
make the same decision? $104,000p + $2,400 = $40,000
o How much could it change
$104,000p = $37,600
before we would make a
different decision? p = $37,600/$104,000 = 0.36

30
31