Certainty Assumption: Decision Analysis Decision Analysis
Certainty Assumption: Decision Analysis Decision Analysis
Certainty assumption
Decision making when the consequences of
alternative decisions are known with a reasonable
Decision Analysis
Decision Analysis degree of certainty
Optimization model: OF specify the estimated
1. Decision analysis under uncertainty consequences of any combination of decisions
(Alternatives!)
2. Decision analysis under Risk
If consequences cannot be predicted with complete
3. EMV certainty, sensitivity analysis
t i t iti it l i
4. Decision trees Decisions often must be made in environments with
5. Utility theory uncertainty
Uncertainty Some Examples
1. A contractor bidding on a new contract.
Decision: carry an umbrella What will be the actual costs of the project?
or not
or not Which other companies might be bidding?
Return or Outcome : get What are their likely bids?
wet or not
2. An agricultural firm selecting the mix of crops and
The outcome depends on
livestock for the upcoming season.
what action nature takes
What will be the weather conditions?
Where are prices headed? What will costs be?
The returns accrue only to
the decision maker. 3. An oil company deciding whether to drill for oil in a
Nature does not care what particular location.
How likely is oil there? How much?
the outcome is
How deep will they need to drill?
Should geologists investigate the site further before drilling?
3 4
1
3/19/2017
Components of Decision Making Payoff table
An individual decision–maker (DM)
Data for decision analysis problem can be represented in
A 'States of Nature' (a set of possible scenarios) “A
a payoff table
a payoff table
state of nature is an actual event that may occur in Payoff table for the decision analysis
the future”
A finite number of possible decision alternatives
(i.e., actions) is available to the DM. Only one action
may be taken.
The return of a decision, called Payoff. “A payoff table
is a means of organizing a decision situation, including
the payoffs from different decisions, given the various Choosing a specific decision depends upon the DM either
states of nature” if he/she is a risk averse or a risk taker
Decision making according to an appropriate criterion
5 6
Cases Cases
Decision are made under four types of environment 3. Under condition of risk:
1. Under condition of certainty: More than one state of nature exist
only one state of nature exist for each alternative DM has sufficient knowledge to allow him assign
probabilities to various state of nature
there is a complete certainty about the future:
4. Under condition of conflict/ partial uncertainty
choose alternatives with max. payoff
Two (or more) opponents with conflicting objectives try to
2. Under condition of uncertainty make decisions with each trying to gain at the cost of the
More than one state of nature exist other(s).
DM lacks sufficient knowledge to allow him assign
DM l k ffi i t k l d t ll hi i Two players struggling to win chess, firms struggling to
Two players struggling to win chess firms struggling to
probabilities to various state of nature maintain market shares, two enemies planning war tactics
Decision maker is working against an intelligent opponent
Theory of games
7 8
2
3/19/2017
Decision making under uncertainty Example / uncertainty
DM has absolutely no knowledge, not even about the likelihood of Company X owns a land that may contain oil.
occurrence for any state of nature
Because of this an oil company has offered to purchase the
In such situations, the decision‐maker's behavior is purely based on
I h it ti th d i i k ' b h i i l b d
his/her attitude toward the unknown:
land for $90,000.
Some Criteria : However, Company X is considering holding the land in
Criterion of pessimism, or conservative: (Maximin or Minmax) order to drill for oil itself. If oil is found, the company’s
Earn no less (or pay no more) than some specified amount expected profit will be approximately $700,000. A loss of
Criterion of optimism, or aggressive (MaxiMax or MiniMin) $100,000 will be incurred if the land is dry (no oil).
Not to miss opportunity to achieve the largest possible profit
pp y g p p Decision of whether to drill or sell based just on these data
(maximax) or lowest possible cost (minimin)
Minimum Regret: (Savag's Opportunity Loss) Decision State of nature
Oil Dry
Drill for oil $700,000 ‐$100,000
Sell the land $90,000 $90,000
9 10
11 12
3
3/19/2017
13 14
Risk and Uncertainty Risk components
Uncertainty: a random chance that something Each risk has a probability of occurring and
will happen, with no way to control whether it possible outcomes (impacts) if it does occur
p ( p )
happens
Risk has two primary components:
Risk: an uncertain event or condition that could
A probability (likelihood) of occurrence of that
negatively impact project performance
event
Common definition of risk : ’the threat or
possibility that an action or event will adversely Impact of the event occurring (amount at stake)
or beneficially affect an organization
or beneficially affect an organization’ss ability to
ability to
achieve its objectives.’ (many other defn.) Risk = f(likelihood, impact)
15 16
4
3/19/2017
Decision making under Risk Example/ Risk
Decision making under risk Company X owns a land that may contain oil.
A probabilistic decision situation A geologist has reported to management that he/she
g g p g /
The decision maker has sufficient information to assign believes there is a chance of 25% of oil.
probability values to the likely occurrence of each of the Because of this an oil company has offered to
various possible states
purchase the land for $90,000.
Criteria for evaluating various alternatives under risk:
However, Company X is considering holding the land
Maximum likelihood criterion
in order to drill for oil itself. If oil is found, the
Expected Monetary Value (EMV) or Expected Return:
Expected Monetary Value (EMV) or Expected Return: company’s expected profit will be approximately
Expected Opportunity Loss $700,000. A loss of $100,000 will be incurred if the
land is dry (no oil).
17 18
Payoff table for the Company X Maximum Likelihood Criterion
Decision State of nature Maximum likelihood criterion:
Oil
Oil Dry
Dry Identify the most likely state of nature (the one with the
largest prior probability).
Drill for oil $700,000 ‐$100,000
For this state of nature, find the action with the
Sell the land $90,000 $90,000 maximum payoff.
Chance of state 0.25 0.75 Choose this action.
Applying to our example:
The Dry state has the largest prior probability. In the Dry
The Dry state has the largest prior probability In the Dry
The given probabilities are prior probabilities
column, the sell alternative has the maximum payoff, so
the choice is to sell the land.
Prior probabilities are subjective (experience or intuition)
19 20
5
3/19/2017
Maximum Likelihood Criterion The Expected Return: Bayes’ decision rule
Rationale: the most important state of nature is the Expected Monetary Value ( EMV)
most likely one, so the action chosen is the best one Most common /rational approach
Using the best available estimates of the probabilities of
U i th b t il bl ti t f th b biliti f
for this particularly important state of nature. the respective states of nature, calculate the expected
Drawback: it completely ignores much relevant value of the payoff for each of the possible actions. Choose
information. the action with the maximum expected payoff.
No state of nature is considered other than the most
likely one. (In a problem with many possible states of
nature, the probability of the most likely one may be
nature the probability of the most likely one may be Where m = number of possible alternatives
quite small) pi = probability of occurrence of alternative i
pij = payoff associated with state of nature Ni and alternative
Sj
21 22
The Expected Return The Expected Return
For the example: Advantage: it incorporates all information, the
E[Payoff (drill)] = 0.25(700) + 0.75(‐100)= 100 payoffs and the best available estimates of states of
E[Payoff (sell)] = 0.25(90) + 0.75(90) = 90 nature (prob.)
Since 100 is larger than 90, the alternative action Drawback: There is no accurate way of predicting the
selected is to drill for oil. future, including a future state of nature
Note that this choice contrasts with the two preceding Under many circumstances, past experience and current
criteria. evidence enable one to develop reasonable estimates of
the probabilities
the probabilities
23 24
6
3/19/2017
The Expected Opportunity loss Sensitivity analysis
An alternative approach to maximizing EMV is to To assess the effect of possible inaccuracies in the
minimize expected opportunity loss (EOL) prior probabilities, it often is helpful to conduct
EOL (or expected value of regrets) sensitivity analysis
The course of action that minimizes the losses or Sensitivity analysis: study the effect of some of the
reduction is the optimal decision alternative. parameters in the mathematical model
Procedure: subtracting the payoff from the maximum Example:
payoff for each event and calculate EOL for each The true prior probability of having oil is likely to be in
alternative and select the min among the alternatives
alternative and select the min among the alternatives th
the range from 0.15 to 0.35, so the corresponding prior
f 0 15 t 0 35 th di i
probability of the land being dry would range from 0.85
to 0.65 (sum =1.0)
25 26
Sensitivity analysis Sensitivity analysis
Evaluate Expected Payoff (EP): Further analysis
Prior probability of oil at the lower end (0.15)
If p = prior probability of oil,
If p = prior probability of oil
E[Payoff (drill)] = 20
The expected payoff from drilling for any p is
E[Payoff (sell)] = 90 Thus, EV =90
When it is at the upper end (0.35) E[Payoff (drill)]=700*p – 100*(1 ‐ p)= 800*p‐
E[Payoff (drill)] = 180 Thus, EV = 180 100
E[Payoff (sell)] = 90 plot of this expected payoff versus p,
The decision swings over
The decision swings over The payoff from selling the land would be 90 for
The payoff from selling the land would be 90 for
The decision is very sensitive to the prior probability any p
27 28
7
3/19/2017
Sensitivity analysis Sensitivity analysis
EP for each alternative action when the prior probability of oil Because the decision Company x should make
changes depends so critically on the true probability of oil,
crossover point
crossover point
serious consideration should be given to conducting
an investigation
E[Payoff (drill)] = E[Payoff
(sell)] Before deciding whether to drill or sell, another option
800p – 100 = 90 is to conduct a detailed seismic survey of the land to
p = 0.2375
Conclusion:
obtain a better estimate of the probability of finding
Sh ld ll h l d if
Should sell the land if oil
oil.
p <0.2375 A case of decision making with experimentation :
Should drill for oil if necessary additional data will be provided.
p > 0.2375
29 30
Decision making under Risk Decision making with experimentation
Decision making between the cases of : Additional testing (experimentation) : improve the
without experimentation (make the decision
without experimentation (make the decision preliminary estimates of the prior probabilities
immediately) Improved estimates are called posterior probabilities.
with experimentation (first do some testing at In our example, conduct a detailed seismic survey of
some expense to reduce the level of uncertainty the land to obtain a better estimate of the probability
about the outcome of the decision) of oil.
The cost is $30,000
31 32
8
3/19/2017
Posterior Probabilities Posterior Probabilities
A seismic survey indicate whether the geological structure In general terms,
is favorable to the presence of oil.
n = number of possible states of nature;
The possible findings of the survey (two states of nature):
Th ibl fi di f h ( f )
P(State = state i) prior probability that true state of
Favorable Seismic Surrounding , FSS (oil is fairly likely);
nature is state i, for i 1, 2, . . . , n;
Unfavorable Seismic Surrounding USS (oil is fairly unlikely)
Finding = finding from experimentation (a random
From past experience:
variable);
if there is oil such seismic study were encouraging (favorable)
60% of the time and they were discouraging (unfavorable) Finding j = one possible value of finding;
40% of the time. P(State = state i/Finding = finding j) = posterior
P(State = state i/Finding = finding j) = posterior
If there is no oil such seismic study were encouraging probability that true state of nature is state i, given that
(favorable) 20% of the time and they were discouraging
(unfavorable) 80% of the time
Finding finding j, for i 1, 2, . . . , n.
33 34
Bayes’ theorem Apply Bayes’ theorem to example
Standard formula of probability theory: If the finding of the seismic survey is unfavorable seismic
soundings (USS), then the posterior probabilities are:
Similarly, if the seismic survey gives favorable seismic
Therefore, for each i
Therefore, for each i 1, 2, . . . , n, the desired formula for the
1, 2, . . . , n, the desired formula for the soundings (FSS), then
di (FSS) th
corresponding posterior probability is :
35 36
9
3/19/2017
Probability tree diagram Posterior probabilities
Calculation
organized Bayes’ decision rule using posterior probabilities:
Input data Col1 *Col2 Cal. PP column‐wise Expected payoffs if finding is unfavorable seismic
soundings (USS):
E[Payoff (drill/Finding= USS)] =1/7(700)+6/7(‐100) ‐ 30 = ‐15.7
E[Payoff (sell/Finding= USS)] =1/7(90) +6/7 (90) ‐ 30 = 60
Expected payoffs if finding is favorable seismic soundings
(FSS):
E[Payoff (drill/Finding= FSS)] =1/2(700) + 1/2(‐100) ‐ 30 = 270
E[Payoff (sell/Finding= FSS)] =1/2 (90) +1/2 (90) ‐ 30 = 60
Subtracting the cost of the experimentation
37 38
Optimal policy Value of experimentation
The optimal policy with experimentation: One method: Expected Value of Experimentation (EVE)
Finding from
g Optimal
p Expected Payoff
p y Expected Payoff
p y EVE = expected payoff with experimentation ‐ expected
Seismic Action Including Cost of excluding Cost of payoff without experimentation
Survey Survey Survey E(Payoff/Finding = USS)= 90
E(Payoff/Finding = FSS) =300 already calculated
USS Sell the 60 90
P(USS) = 0.7, P(FSS) = 0.3
land
FSS Drill for oil 270 300 Expected payoff with experimentation = 0.7(90) +
( )
0.3(300) = 153
Value of Experimentation Thus, EVE = 153 ‐ 100
The analysis does not answer : whether it is worth spending Since this value exceeds 30, the cost of conducting a
$30,000 to conduct the experimentation (the seismic survey)
detailed seismic survey should be done
39 40
10
3/19/2017
Decision trees
Helpful when a sequence of decisions must be made
Decision Trees Decision trees useful:
visually displaying the problem
Organizing the computational work
Our example involves a sequence of two decisions:
1. Should a seismic survey be conducted before an
action is chosen?
2. Which action (drill for oil or sell the land) should be
chosen?
Draw the tree
41 42
Decision trees Decision trees
The decision tree (before including any numbers) example Symbols:
Nodes ( forks) and
Arcs (branches)
A decision fork : represented by a square , a decision
needs to be made
A chance fork, represented by a circle, a random event
occurs at that point
The path followed from start (fork a ) to any terminal
The path followed from start (fork a ) to any terminal
branch is determined both by the decisions made and
by random events
43 44
11
3/19/2017
Decision tree with numbers
Decision trees
Numbers in the tree:
Numbers under or over the branches : cash flows
Probability are in parentheses along branches
For each path from node a to a terminal branch, the
numbers are added to obtain the resulting total payoff
For chance fork h, the probabilities are prior
probabilities
For chance fork
For chance fork f, g, the probabilities are posterior
f g the probabilities are posterior
probabilities, given the finding from the seismic survey
45 46
Decision trees Decision trees
Analysis steps: For chance forks f, g, and h :
1. Start at the right side of the decision tree and move
left one column at a time.
2. For each chance fork, calculate its expected payoff
Record this expected payoff for each fork in boldface
next to the fork
3. For each decision fork, compare the expected payoffs For decision forks c, d, and e
of its branches and choose the alternative whose branch Fork c: Drill alternative has EP
Fork c: Drill alternative has EP = ‐15.7
15.7
has the largest expected payoff Sell alternative has EP = 60
4. In each case, record the choice on the decision tree 60 > ‐ 15.7, so choose the Sell alternative
by inserting a double dash as a barrier through each Do the same for fork d and e
rejected branch
47 48
12
3/19/2017
Decision trees Decision trees
For fork b :
EP = 0.7(60) + 0.3(270) = 123,
Finally, fork a
Do seismic survey has EP = 123 The final decision
No seismic survey has EP = 100 tree that records
the analysis
123 > 100, so choose: Do seismic survey
49 50
Decision trees Software: Treeplan
Following the open paths from left to right the TreePlan is a software, an excel add‐in to build decision trees in Excel.
It was developed by Prof. Michael Middleton, Univ. SF, California (info
Optimal policy: and demo version: www treeplan com ) Free: academic version
and demo version: www.treeplan.com ) Free: academic version
Do the seismic survey
If the result is unfavorable, sell the land
If the result is favorable, drill for oil
The expected payoff (including the cost of the seismic
survey) is 123 ($123,000).
Backward induction procedure
Backward induction procedure
51 52
13
3/19/2017
Utility theory
Assumption : expected payoff in monetary terms is
Utility theory the appropriate measure for DA.
Assumption appropriate?
Some examples:
Ismamalehu/Alsmamam?
(1) Accepting a 50:50 chance of winning 10,000 Birr or 5 Birr
(2) receiving Birr 4,000 with certainty (Bank)
A company not willing to invest in a new product even
when the expected profit is substantial if there is a risk
h th t d fit i b t ti l if th i ik
of losing its investment/bankrupt
Contrary to Bayes’ decision rule! Paradox: Utility
theory !
53 54
Example 1 : investment problem The Meaning of Utility
Payoff table
Utilities are used when the decision criteria must be based
on more than just expected monetary values
on more than just expected monetary values.
Decision State of Nature
Utility is a measure of the total worth of a particular
Alternatives Prices Up, S1 Prices Stable, S2 Prices Down, S3 outcome, reflecting the decision maker’s attitude towards a
Investment A, d1 $30,000 $20,000 ‐50,000 collection of factors. (Some of these factors may be profit,
loss, and risk.)
Investment B, d2 50,000 ‐20,000 ‐30,000
Define a utility function that would reflect individual’s
Investment C, d3 0 0 0 preference
P(S1) = 0.3
)=03 P(S2) = 0.5
)=05 P(S3)
) = 0.2
0.2 Utility theory is based on rationality and maximization of
Utility theory is based on rationality and maximization of
utility
EV(d1) = $9,000 EV(d2) = ‐$1,000 EV(d3) = $0 Maximizing someone’s utility automatically determines
his/her preferred options
Alt. 1 (d1) is the rational decision
In economics, ‘Utility’ means the satisfaction obtained from
Rationality assumption: humans always seek the best alternative consuming a commodity.
55 in a set of possible choices 56
14
3/19/2017
Utility function Utility function
Utility determination is subjective (depends on
A typical utility function for money
attitude to risk) Utility function (UF)
Utility function (UF)
Utility theory approach determine utility function (UF) may not be a linear
for a decision maker function of income
UF converts money into an arbitrary utility measure or wealth
(Each $ is not equally valuable to an individual) Note in the graph:
Utility $30,000 is
only twice as much
only twice as much
as the utility of
$10,000
57 58
Utility theory Utility Functions
Attitude toward risk:
Utility
Risk‐averse : (the above graph): a concave utility function
A decreasing marginal utility for money: decreasing slope of
A decreasing marginal utility for money: decreasing slope of 100
the function as the amount of money increases. Risk avoider
Individuals purchasing insurance exhibit risk avoidance 80
behavior
Risk seekers : function is convex 60
The slope of their utility function increases as the amount of
money increases, so they have an increasing marginal utility 40
for money.
Risk‐neutral : 20 Risk taker
individual’s utility for money is simply proportional to the
amount of money involved
expected utility is equal to expected return -60 -40 -20 0 20 40 60 80 100
Monetary Value (in $1000’s)
59 60
15
3/19/2017
Steps for Determining the Utility of Money
Utility theory
Attitude toward risk
Step 1:
exhibit a mixture of these kinds of behavior
Develop a payoff table using monetary values
Develop a payoff table using monetary values.
can shift over time
can vary when dealing with one’s personal finances and
an organization Step 2:
UF for money has an important implication for Identify the best and worst payoff values and assign
decision making in the face of uncertainty each a utility value, with
When a UF for money is incorporated into a decision U(best payoff) > U(worst payoff).
( p y ) ( p y )
analysis approach to a problem, it must be
Example 1:
constructed to fit the preferences and values of the
decision maker Best payoff = 50,000
Worst payoff =‐50,000
61 62
Steps for Determining the Utility of Money
Example 1 : investment problem
Step 3:
Payoff table For every other monetary value M in the payoff table:
3a: Define the lottery. The best payoff is obtained
3a: Define the lottery. The best payoff is obtained
Decision State of Nature with probability p; the worst is obtained with
Alternatives Prices Up, S1 Prices Stable, S2 Prices Down, S3 probability (1 – p).
Investment A, d1 $30,000 $20,000 ‐50,000
Investment B, d2 50,000 ‐20,000 ‐30,000
Investment C, d3 0 0 0
Lottery: obtains a payoff of $50,000 with
P(S1)
) = 0.3
0.3
probability of p and a payoff ‐$50,000
P(S2) = 0.5
)=05 P(S3) = 0.2
)=02
with probability of (1‐p)
EV(d1) = $9,000 EV(d2) = ‐$1,000 EV(d3) = $0
63 64
16
3/19/2017
Steps for Determining the Utility of Money
Example : step 3a 3b: Determine the value of p such that the decision
Consider a gambling situation in which an individual is maker is indifferent between a guaranteed
offered two alternatives involving three outcomes: $ payoff of M and the lottery defined in step 3a.
5,000, $2,500, $0 3c: Calculate the utility of M:
(i) guaranteed cash of $2,500 U(M) = pU(best payoff) + (1 – p)U(worst payoff)
(iI) winning $5,000 with p=0.5 and $0 with(1‐P)=0.5
Assume she/he opts for (i), decrease (i) to $2,000, Step 4:
decrease (i) until the individual is indifferent to the two Convert the payoff table from monetary values to utility
alternatives ( say $1,000). This point is called the values.
certainty monetary equivalent (CME) of the gamble.
Utility of $1,000 with p=1.0 and utility of $5,000 with
Step 5:
p=0.5 and nothing with p=0.5 becomes equal
Apply the expected utility approach to the utility table
developed in step 4, and select the decision alternative with
65 66 the highest expected utility.
Expected Utility Approach Example 2
The following UF is given for the DM in Example 1
Once a utility function has been determined, the
optimal decision can be chosen using the expected
ti l d i i b h i th t d UF of the DM
utility approach. M U(M) 12
‐50 0 10
The utility corresponding to each state of nature is
‐30 4
multiplied by the probability for that state of nature. 8
Utility
‐20 5.5
6
The sum of these products for each decision alternative 0 7.5
4
represents the expected utility for that alternative.
represents the expected utility for that alternative. 20 9
30 9.5 2
The decision alternative with the highest expected 0
50 10
utility is chosen. ‐50 ‐30 ‐20 0 20 30 50
Money in thousands of Birr
67 68
17
3/19/2017
Applying Utility Theory to Company x
Example 2: investment problem example
Decision State of Nature Company X was operating without much capital, so a
Alternatives Prices Up, S1 Prices Stable, S2 Prices Down, S3 loss of $100,000 would be quite serious (already has
Investment A, d1 $30,000 $20,000 ‐50,000 gone heavily into debt to keep going)
Investment B, d2 50,000 ‐20,000 ‐30,000
The worst‐case scenario would be to come up with
Investment C, d3 0 0 0
$30,000 for a seismic survey and then still lose
P(S1) = 0.3 P(S2) = 0.5 P(S3) = 0.2
$100,000 by drilling when there is no oil (bankrupt !)
Utility table.
On the other hand, striking oil is an exciting
Decision State of Nature
Alternatives prospect, earning $700,000 (solid financial footing)
t i $700 000 ( lid fi i l f ti )
Prices Up, S1 Prices Stable, S2 Prices Down, S3
Investment A, d1 9.5 9 0
Investment B, d2 10 5.5 4
Investment C, d3 7.5 7.5 7.5
69 EU(d1) = 7.35 EU(d2) = 6.55 EU(d3) = 7.5 70
Applying Utility Theory Note:
u(M) essentially equals
Apply the Expected Utility Approach
M for small values
Suppose the following is given as UF of the DM:
pp g g
u(M) gradually falls off
(M) d ll f ll ff
Monetary utility M for larger values of
Payoff
M.
‐130 ‐150
This is typical for a
‐100 ‐105
moderately risk‐averse
60 60
90 90
670 580
How are these values obtained ? Read lecture
700 600
notes
71 72
18
3/19/2017
Exercise (class work)
Apply the utility table (utility function) developed Payoff utility
p y
for Company X and determine the rational
decision. ‐130 ‐150
(select the decision alternative with the highest ‐100 ‐105
expected utility)
60 60
90 90
Considering the Utility the best decision is :
C id i h U ili h b d i i i 670 580
No seismic survey and sell the land ???
700 600
73 74
Summary
Decision analysis : an important technique for decision
making in the face of uncertainty.
It requires: enumerating all the available courses of
It requires: enumerating all the available courses of
action, identifying the payoffs for all possible
outcomes, and quantifying the subjective probabilities for
all the possible random events
The analysis can incorporate to perform experimentation
to obtain better estimates of the probabilities of the
possible states of nature
Decision trees are a useful visual tool for analyzing this
Decision trees are a useful visual tool for analyzing this
option or any series of decisions
Utility theory provides a way of incorporating the decision
maker’s attitude toward risk into the analysis
75
19