Management Science

1
Decision Analysis

Module 12 Decision Analysis

At the end of this module you are expected to:

1. Discover the components of decision making
2. Discuss decision making in situations of certainty
3. Explain decision making in scenarios of uncertainty

In previous discussions, we talked about linear programming, and the creation of

various models to solve problems, and in the aim of helping owners or managers in day to
day decision making situations. Values for decision variables were used as representations
of the solutions to the models. However, these solutions were in an assumed situation
where the assumptions wherein certainty is present. Meaning, the linear programming
model understood all given coefficients, values of constraints, and values for solution did
not vary and were known with certainty.

In reality or actual application, most situations involved in decision-making happen

in conditions of uncertainty. To give an sample situation would be the demand of a certain
product, which may not be an order of 100 units next week, instead it will be 200 order
units or maybe 50 order units, and it all depend on the market’s state which is for all
intents and purposes uncertain.

There are two classifications of decision situations---and these are: scenarios where
probabilities of future situations that cannot be assigned (decision making under certainty)
and situations where probabilities can be assigned (decision making under uncertainty).
Before moving on to the discussion of decisions classes and the decision-criterion that is
associated with them, let us first discuss decision making components.

A. Decision-Making Components

States of nature are what we call the actual decision itself and the events in actuality
that could happen in the future. These are but a couple of the components of a decision-

Course Module
 Management Science
2
Decision Analysis

making situation. At the point in time that a decision is made, it is not certain to the
decision-maker what state of nature will happen in the future nor that he/she would have
the capability to control them.

Let us say that a distribution business is thinking on buying a computer with and
has a goal to increase the amount of orders it is processing which would result in an
increase in business. We can therefore assume that the company may experience a
significant increase in profit, granted that the current economic will continue its good
conditions and remain stable; but the company will lose money if the economy takes a
downturn. In this example, the situations where we need to make a decision, or the
possible solutions are to buy or not to buy the computer. We can think of the possible
states of nature in this situation; first it that it can be on economic conditions that is good
and economic conditions that is not good (bad). Decision outcome is determined by the
state of nature that occurs. Obviously, the decision maker has no way of controlling what
state of nature will happen.

To better analyze these types of decision-situations and to come up with the best
decisions, we must organize them into payoff tables. Because of the different states of
nature in a decision problem, there is a need to make a payoff table to organize and
illustrate the different decision’s payoffs.

Table 1 Payoff Table

For a certain state of nature that may happen in the future as shown in Table 1, each
decision will lead in an outcome that is referred to as a “payoff”. Payoffs can be illustrated
 Management Science
3
Decision Analysis

by means of using different kinds of values but payoffs are often illustrated by means of
cost or profit revenues.

To clearly understand the components of decision making the following list will be
of good use:

• A decision maker responsible for making decisions is needed by a decision

• Although there are many alternatives to choose from, a decision maker may
only choose one.
• The objective of the decision maker is to make a choice which among the
alternatives presented is the best.
• Once the best alternative is chosen and the decision is made, the succeeding
events that will happen will not be in the hands of the decision-maker.
• Outcomes with measurable values arise from each chosen alternative

B. Decision Making Under Certainty

This basically is where we know with certainty which event will occur. Our
consideration is only one event and our method of solution would be obvious. We evaluate
the outcome of the alternatives in play and then eventually choosing which among the
alternative will produce the best outcome possible.

Suppose that we have P10,000 to invest in a year and we would like to decide on
which investment will give us the best return in a year. Obviously simplified but this is
basically how we arrive at a decision in this case:

Alternatives Event (Get Principal with interest at year end)

Bank P10,600
Government Stock P10,900
Stock Market P10,800

There is only one event, that is to invest and get an interest or return by year end. It
is certain that there will be an interest gained by year end. Each of the alternatives also
Course Module
 Management Science
4
Decision Analysis

possess values which are certain by the year end. Therefore, what is left to do is to decide
which of the alternatives will give the best result. And as shown in our example, the best
result would be to invest in a Government Stock showing the greatest amount of P10,900.

C. Decision Making Under Uncertainty

Let us start with an example on developing a payoff table that is without

probabilities. Let us discuss this example: An investor is weighing a decision on purchasing
one of these real estate types: a warehouse, an apartment building, and a office building. To
determine the amount of profit the investor will make will be depending on the future
states of nature which can either economic conditions that are good or economic
conditions that are bad or poor.

The resulting profits in every possible decision situation made in each identified
economic condition is depicted in the payoff table below:

Table 2 Payoff Table for the real estate investment

After organizing and creating the payoff table, the actual decision will now be based
on several decision-making criteria. In the end, the decision-maker will have to choose
which among the identified criterion or the best combination would suit his/her need best.

The Maximax Criterion

 Management Science
5
Decision Analysis

It pertains to the maximum payoffs that one can get that is of the highest
level or the maximum of the maximum (in terms of payoffs), and this is obviously
where the criterion got the name “maximax”. The decision maker will select the
decision that would result in the maximum of the maximum payoffs. This type of
criterion is said to lean on the optimistic side. The most encouraging state of nature
for every alternative decision is in a situation where it is assumed by the decision
maker to occur. In this scenario, when one chooses this criterion, our investor
example would have an optimistic assumption that there would be continuous good
economic conditions that will occur in the future.

The maximax criterion when it is used by the investor in the following table
shows each decision’s maximum payoff is first selected and then the maximum out
of all of them is then selected.

Table 3 Payoff Table illustrating Maximax decision

Out of the three maximum payoffs can be $50,000, $100,000, and $30,000 -
it is shown here that $100,000 is the maximum. Therefore, it would be the best the
decision in this situation to buy the office building.

Note that even though this decision results in the highest payoff of $100,000,
it completely ignores the possibility of a potential loss of -$40,000. Using a maximax
criterion, the decision maker assumes a very positive gain in the future with respect
to the state of nature.

Course Module
Management Science
6
Decision Analysis

Another important thing to note here is that this example is focused on profit.
IN a scenario where the payoff table consists of costs, then the opposite of these
options will be illustrated. It will then deal with the minimum of the minimum costs,
or it would be called the Minimin Criterion. The same logic will be applied but this
time in will be in terms of costs.

The Maximin Criterion

The maximin criterion is pessimistic which is the opposite idea with the
maximax criterion that is optimistic. In this criterion, the decision maker selects the
maximum of the minimum payoffs. There is an assumption on the part of the
decision-maker that for each decision alternative, the minimum payoff happens. The
maximum is chosen from these minimum payoffs. This is illustrated in Table 4.

Table 4 Payoff Table illustrating Maximin decision

As shown in Table 4, the minimum payoffs are $30,000, -$40,000, and
$10,000. $30,000 is the maximum of these payoffs. Therefore, using this criterion
the decision in buying the apartment building is the logical choice. This may seem as
a conservative decision because the choices that were taken inconsideration are
only the bad result that may happen. In the maximax criterion, the decision to buy
the office building included the possibility of a large loss -$40,000. The choice to buy
the apartment building at worst gives a profit of $30,000. But the largest biggest
probable gain from the apartment purchase is lesser than that of buying the office
 Management Science
7
Decision Analysis

building that is $50,000 versus $100,000. And again, if the table consists of costs
rather than profits, then the conservative approach is to select the maximum cost
for each decision. Then, in the decision which results in the minimum, one selects
the minimax of these costs.

The Minimax Regret Criterion

Let us say that the investor would make the decision to buy the warehouse,
but then discovered that future economic conditions were a lot better than
projected. It is then a natural reaction for the investor to be disappointed for not
purchasing the office building that may have resulted in the maximum payoff of
$100,000 which is on a good economic condition.

The investor may be disappointed about the decision made in purchasing the
warehouse. The level of disappointment is the monetary value of $70,000 which is
the difference payoff choice of the investor compared to the best choice. Given this
case, $100,000 is the maximum payoff that the investor can get in good economic
conditions and $30,000 is the the maximum payoff an investor can get in poor
economic conditions.

When there is an attempt by the decision maker to avoid regret which is

done by choosing the alternativethat would minimize the maximum regret is called
the minimax regret criterion. In using the minimax regret criterion, the decision
maker must first choose the maximum payoffs in every state of nature identified
which as we have already computed $100,000 and $30,000. All added payoffs in
each identified states of nature will be subtracted from these amounts as follows:

Course Module
Management Science
8
Decision Analysis

If a decision results in less than the maximum payoff, then the decision
maker would experience regret as represented by the values computed above.
These values are summed up in a customized adaptation of the payoff table referred
as a regret table. It is also at times called a table for opportunity loss which makes
the term opportunity loss tantamount to the feeling of regret.

Table 5 Regret Table

After determining the maximum regret for each decision, to use the minimax
regret criterion the choice that corresponds to the minimum of these regret values
is eventually selected.

Table 6 Regret Table using Minimax Regret Decision

The philosophy of the minimax regret criterion aims to achieve for the
investor to have the slightest amount of regret possible. In our example, according
to this criterion, purchasing the apartment building is the decision relative to the
 Management Science
9
Decision Analysis

warehouse purchase or office building purchase. In other words, or in monetary

terms, both the office building and warehouse purchase will result in a $70,000
regret as opposed to the $50,000 regret of the apartment building.

The Hurwicz Criterion

This type of criterion strikes a middle ground between the maximax and
maximin criteria and this is calle the Hurwicz Criterion. The standard of this
criterion in making decision is that it is neither totally pessimistic or optimistic. In
this criterion, when decision payoffs are weighted, it is by means of coefficient of
optimism, this refers to the measure of optimism of the decision maker. This
coefficient of optimism, identified as , which is between zero and one. If  equals
to 0, this would signal the complete pessimism of the decision maker. If  equals to
1, this means that the the decision maker is optimistic in a complete manner. With
this illustration, we can say that if  is the coefficient of optimism, then 1- is the
coefficient of pessimism.

This decision criterion entails that for every alternative decision, the
maximum payoff should be multiplied by  and the minimum payoff multiplied by
1-. In the above mention scenario of the investor,, if  = .4 then 1 - .4 = .6 resulting
to the following values:

This criterion indicates the range of the decision alternative that corresponds
to $38,000 which is the maximum weighted value in the example mentioned above.
The Hurwicz criterion is the maximin criterion when  = 0. When  = 1, this will
now be the maximax criterion. The restriction of the Hurwicz criterion is that
decision maker must determine the . And it most certainly would be a challenge for

Course Module
 Management Science
10
Decision Analysis

the decision maker to make a determination with regard to the degree of his/her
optimism.

And since it is a completely subjective measure, we can say that this is a

decision criterion that is completely subjective.

The Equal Likelihood Criterion

This criterion assumes that the states of nature that would likely ohappen
equally is called the the LaPlace or Equal Likelihood Criterion which happens
decision payoff for every state of nature by an equal weight is multiplied.

Now since our example has 2 states of nature, a .50 weight is assigned to each
one. We then multiply each weight by every payoff in every identified decision:

Much like the Hurwicz criterion, the decision to be made should be the one
that entails the maximum of these weighted values. With $40,000 being the highest
value, the decision of the investor is to buy the apartment building.

We have placed this criterion under decision making under certainty despite
the 50% probability just because of the fact that it is still certain that one of the
states is likely to occur. We are assuming a .50 probability or a 50% chance that
either state of nature will occur by applying the equal likelihood criterion. This same
basic logic can be used in giving weights to states of nature unequally or differently
in many decision problems. It is another way of saying that different probabilities
can be assigned to each state of nature, where it can be indicated that one state is
more likely to occur than the other.

Ultimately, the use of which criterion or collection of criteria and the final decision
would depend on the decision maker’s characteristics and philosophy. If he or she is
 Management Science
11
Decision Analysis

extremely optimistic, then the decision maker might steer clear of other results and decide
to buy the office building due to the maximax criterion that is quite identical his or her
personal or subjective choice and philosophy in decision-making .

Given the possibilities that one can be assign, the decision maker has several
criterion to help him or her in making decisions. What we will discuss here is the projected
value and the anticipated opportunity loss.

Expected Value

First, the decision maker must make an estimation of the the possibility of
each state of nature’s occurrence in order to apply the anticipated value as a
criterion in decision-making. After the estimates are done, the projected value for
every decision alternative may then be computed.

To compute the expected value, multiply every possible outcome of a

decision made by the probability of the rate of occurence and then adding the
products. The EV or expected value of a arbitrary variable referred as x can be
symbolically written as EV(x), and the computation will be:

Going back to the example scenario of the real estate, and assume that,
basing on different economic forecasts, the investor approximates a .60 probability
that the economic conditions will be good and a .40 probability that economic
conditions will be poor. This latest data is illustrated in the following table:

Course Module
Management Science
12
Decision Analysis

Table 7 Payoff Table with Probabilities for states of nature

The most decision is obviously the choice that will bring $44,000 which is the
highest expected value. Therefore, the best decision is buying the office building. Do
note that this does imply that the investment will result to $44,000, but rather it is
one can assume that one of the payoff values ($100,000 or -$40,000) will be the
result. The expected value implies that if such situation on making decision
occurred, in multiple number of times, $44,000 will be the average payoff outcome.
Then again, if the payoffs were in conditions of costs, then the lowest expected value
is the best decision.

Expected Opportunity Loss

Closely related to the expected value is a decision criterion called expected

opportunity loss. To use this decision criterion, one must multiply the probabilities
by the regret or opportunity loss of every decision outcome instead of multiplying
the decision outcomes by the probability of them happening.
 Management Science
13
Decision Analysis

In the minimax regret criterion, we encountered the concept of regret. Using

the data from Table 6 Regret Table and incorporating the probabilities of happening
of every state of nature that may occur, we come up with the following table:

Much like the minimax regret criterion, minimizing the regret results in the
best solution. For this example, minimizing the expected regret or opportunity loss
results in the best solution. $28,000 is the expected regret at a minimum, therefore,
best option is to buy the office building property.
If you will notice, the recommended decision by both the expected value and
the expected opportunity loss are the similar, and that is to buy the building of the
office property. Note that this is no coincidence; these two techniques often end in

Course Module
Management Science
14
Decision Analysis

the same decision. It therefore means that it will be redundant if we apply the two
methods to come up with a decision when one of them will suffice.
Again, very important to note is that the decisions made either from expected
opportinuty loss criteria and the expected value criteria will depend completely on
the estimates of probability of the decision-maker. This means that when one uses
probabilities that are not accurate, it will lead to errors on the decisions made. The
decision maker as much as possible must ensure accuracy in the determination of
the probability of every state of nature.