Modeling

INTRODUCTION
Management science is a discipline that attempts to aid managerial decision making by
applying a scientific approach to managerial problems that involve quantitative factors.

Management Science Uses a Scientific Approach:

Step 1: Define the problem and gather data.
Step 2: Formulate a model (typically a mathematical model) to represent the problem.
Step 3: Develop a computer-based procedure for deriving solutions to the problem from
the model.
Step 4: Test the model and refine it as needed.
Step 5: Apply the model to analyze the problem and develop recommendations for
management.
Step 6: Help to implement the team’s recommendations that are adopted by management.

CHAPTER 2
I/ Three questions need to be answered to begin the process of using the spreadsheet
to formulate a mathematical model (in this case, a linear programming model) for the
problem.
1/ WHAT WE ARE FINDING/WHAT ARE THE DECISIONS TO BE MADE? => Decision
variables?
2/ WHAT IS OUR OBJECTIVE/ THE OVERALL MEASURE OF PERFORMANCE? =>
Objective function?
3/ WHAT ARE THE CONDITIONS ON THESE DECISIONS? => Constraints?

⇒ MATHEMATICAL MODEL/ ALGEBRAIC MODEL

II/ Characteristics of a Linear Programming Model on a Spreadsheet:

1. Decisions need to be made on the levels of a number of activities, so changing cells
are used to display these levels.
2. These activity levels can have any value (including fractional values) that satisfy a
number of constraints.

3. Each constraint describes a restriction on the feasible values for the levels of the
activities, where a constraint commonly is displayed by having an output cell on the left,
a mathematical sign (≤, ≥, or =) in the middle, and a data cell on the right.
4. The decisions on activity levels are to be based on an overall measure of performance,
which is entered in the objective cell. The goal is to either maximize the objective cell or
minimize the objective cell, depending on the nature of the measure of performance.
5. The Excel equation for each output cell (including the objective cell) can be expressed
as a SUMPRODUCT function,2 where each term in the sum is the product of a data cell
and a changing cell.
Linear programming models are not the only models that can have characteristics 1, 3,
and 4. However, characteristics 2 and 5 are key assumptions of linear programming.
Therefore, these are the two key characteristics that together differentiate a linear
programming model from other kinds of mathematical models that can be formulated on
a spreadsheet.
III/ Summary of the Formulation Procedure:
Here is a summary of the steps involved in the procedure.
1. Gather the data for the problem.
2. Enter the data into data cells on a spreadsheet.
3. Identify the decisions to be made on the levels of activities and designate changing
cells for displaying these decisions.
4. Identify the constraints on these decisions and introduce output cells as needed to
specify these constraints.
5. Choose the overall measure of performance to be entered into the objective cell.
6. Use a SUMPRODUCT function to enter the appropriate value into each output cell
(including the objective cell).

IV/ Formulating the Wyndor Model Algebraically:

1. Gather the relevant data.
2. Identify the decisions to be made.
3. Identify the constraints on these decisions.
4. Identify the overall measure of performance for these decisions.
5. Convert the verbal description of the constraints and measure of performance into
quantitative expressions in terms of the data and decisions.
Eg. Let D = number of doors (number of the special new doors to be produced per week)
W = number of windows (number of the special new windows to be produced per
week)
Choose the values of D and W so as to maximize (P = Profit (total profit per week from
the two products, in dollars)
P = 300D + 500W

subject to satisfying all the following constraints:

D≤4
2W ≤ 12
3D + 2W ≤ 18
and
D≥0W≥0

V/ Terminology for Linear Programming Models:

Much of the terminology of algebraic models also is sometimes used with spreadsheet
models. Here are the key terms for both kinds of models in the context of the Wyndor
problem.
1. D and W are the decision variables.
2. 300D + 500W [or SUMPRODUCT (UnitProfit, UnitsProduced)] is the objective
function.
3. P is the value of the objective function (or objective value for short).
4. D ≥ 0 and W ≥ 0 are called the nonnegativity constraints (or nonnegativity
conditions).
5. The other constraints are referred to as functional constraints (or structural
constraints).
6. The parameters of the model are the constants in the algebraic model (the numbers in
the data cells).
7. Any choice of values for the decision variables (regardless of how desirable or
undesirable the choice) is called a solution for the model.

8. A feasible solution is one that satisfies all the constraints, whereas an infeasible -
solution violates at least one constraint.
9. The best feasible solution, the one that maximizes P (or G12), is called the optimal -
solution. (It is possible to have a tie for the best feasible solution, in which case all the
tied solutions are called optimal solutions.)
=> Interpret the results: 3 zones (decision variables, optimal solution and constraints)

CHAPTER 3
Five key categories of linear programming problems:
I/ Resource-allocation problems:
Resource-allocation problems are linear programming problems involving the allocation
of resources to activities. The identifying feature for any such problem is that each
functional constraint in the linear programming model is a resource constraint, which has
the form
Amount of resource used ≤ Amount of resource available
for one of the resources.
The amount of a resource used depends on which activities are undertaken, the levels of
those activities, and how heavily those activities need to use the resource. Thus, the
resource constraints place limits on the levels of the activities. The objective is to choose
the levels of the activities so as to maximize some overall measure of performance (such
as total profit) from the activities while satisfying all the resource constraints.
Summary of the Formulation Procedure for Resource-Allocation Problems:
The four examples illustrate that the following steps are used for any resource-allocation
problem to define the specific problem, gather the relevant data, and then formulate the
linear programming model.
1. Since any linear programming problem involves finding the best mix of levels of various
activities, identify these activities for the problem at hand. The decisions to be made are
the levels of these activities.
2. From the viewpoint of management, identify an appropriate overall measure of
performance (commonly profit, or a surrogate for profit) for solutions of the problem.
3. For each activity, estimate the contribution per unit of the activity to this overall measure
of performance.

4. Identify the resources that must be allocated to the activities.

5. For each resource, identify the amount available and then the amount used per unit of
each activity.
6. Enter the data gathered in steps 3 and 5 into data cells in a spreadsheet. A convenient
format is to have the data associated with each activity in a separate column, the data for
the unit profit and each constraint in a separate row, and to leave two blank columns
between the activity columns and the amount of resource available column. Figure 3.4
shows a template of the overall format of a spreadsheet model for resource-allocation
problems.
7. Designate changing cells for displaying the decisions on activity levels.
8. For the two blank columns created in step 6, use the left one as a Totals column for
output cells and enter ≤ signs into the right one for all the resources. In the row for each
resource, use the SUMPRODUCT function to enter the total amount used in the Totals
column.
9. Designate an objective cell for displaying the overall measure of performance. Use a
SUMPRODUCT function to enter this measure of performance.
All the functional constraints in this linear programming model in a spreadsheet are
resource constraints, that is, constraints with a ≤ sign. This is the identifying feature that
classifies the problem as being a resource-allocation problem.

II/ Cost-benefit-trade-off problems:

Cost–benefit–trade-off problems have a form that is very different from resource-
allocation problems. The difference arises from managerial objectives that are very
different for the two kinds of problems.

For resource-allocation problems, limits are set on the use of various resources (including
financial resources), and then the objective is to make the most effective use (according
to some overall measure of performance) of these given resources.
For cost–benefit–trade-off problems, management takes a more aggressive stance,
prescribing what benefits must be achieved by the activities under consideration
(regardless of the resulting resource usage), and then the objective is to achieve all these
benefits with minimum cost. By prescribing a minimum acceptable level for each kind of
benefit, and then minimizing the cost needed to achieve these levels, management hopes
to obtain an appropriate trade-off between cost and benefits.
Cost–benefit–trade-off problems are linear programming problems where the mix of
levels of various activities is chosen to achieve minimum acceptable levels for various
benefits at a minimum cost. The identifying feature is that each functional constraint is a
benefit constraint, which has the form
Level achieved ≥ Minimum acceptable level
for one of the benefits.
Interpreting benefit broadly, we can think of any functional constraint with a ≥ sign as a
benefit constraint. In most cases, the minimum acceptable level will be prescribed by
management as a policy decision, but occasionally this number will be dictated by other
circumstances.
For any cost–benefit–trade-off problem, a major part of the study involves identifying all
the activities and benefits that should be considered and then gathering the data relevant
to these activities and benefits.
Three kinds of data are needed:
1. The minimum acceptable level for each benefit (a managerial policy decision).
2. For each benefit, the contribution of each activity to that benefit (per unit of the activity).
3. The cost per unit of each activity.
Summary of the Formulation Procedure for Cost–Benefit–Trade-Off Problems
The nine steps in formulating any cost–benefit–trade-off problem follow the same pattern
as presented at the end of the preceding section for resource-allocation problems, so we
will not repeat them here. The main differences are that the overall measure of
performance now is the total cost of the activities (or some surrogate of total cost chosen
by management) in steps 2 and 3, benefits now replace resources in steps 4 and 5, and
≥ signs now are entered to the right of the output cells for benefits in step 8. Figure 3.6
shows a template of the format of a spreadsheet model for cost–benefit–trade-off
problems.
All the functional constraints in the resulting model are benefit constraints, that is,
constraints with a ≥ sign. This is the identifying feature of a pure cost–benefit–trade-off
problem.

III/ Mixed problems:

Many linear programming problems do not fit completely into any of the previously
discussed categories (pure resource-allocation problems, cost–benefit–trade-off
problems, and fixed-requirement problems) because the problem’s functional constraints
include more than one of the types. Such problems are called mixed problems.

The identifying feature of a pure resource-allocation problem is that all its functional
constraints are resource constraints. The identifying feature of a pure cost–benefit–trade-
off problem is that all its functional constraints are benefit constraints.
The last of the three types of functional constraints is fixed-requirement constraints, which
require that the left-hand side of each such constraint must exactly equal some fixed
amount. Thus, since the left-hand side represents the amount provided of some quantity,
the form of a fixed-requirement constraint is
 Amount provided = Required amount
The identifying feature of a pure fixed-requirements problem is that it is a linear
programming problem where all its functional constraints are fixed-requirement
constraints. The next two sections will describe two particularly prominent types of fixed-
requirement problems called transportation problems and assignment problems.
Summary of the Formulation Procedure for Mixed Linear Programming Problems
The procedure for formulating mixed problems is similar to the one outlined for resource-
allocation problems. However, pure resource-allocation problems only have resource
constraints whereas mixed problems can include all three types of functional constraints
(resource constraints, benefit constraints, and fixed-requirement constraints). Therefore,
the following summary for formulating mixed problems includes separate steps for dealing
with these different types of constraints

1. Since any linear programming problem involves finding the best mix of levels of various
activities, identify these activities for the problem at hand. The decisions to be made are
the levels of these activities.
2. From the viewpoint of management, identify an appropriate overall measure of
performance for solutions of the problem.
3. For each activity, estimate the contribution per unit of the activity to this overall measure
of performance.
4. Identify any resources that must be allocated to the activities. For each one, identify
the amount available and then the amount used per unit of each activity.

5. Identify any benefits to be obtained from the activities. For each one, identify the
minimum acceptable level prescribed by management and then the benefit contribution
per unit of each activity.
6. Identify any fixed requirements that, for some type of quantity, the amount provided
must equal a required amount. For each fixed requirement, identify the required amount
and then the contribution toward this required amount per unit of each activity.
7. Enter the data into data cells in a spreadsheet.
8. Designate changing cells for displaying the decisions on activity levels.
9. Use output cells to specify the constraints on resources, benefits, and fixed
requirements.
10. Designate an objective cell for displaying the overall measure of performance.

IV/ Transportation problems:

One of the most common applications of linear programming involves optimizing a
shipping plan for transporting goods. In a typical application, a company has several
plants producing a certain product that needs to be shipped to the company’s customers
(or perhaps to distribution centers). How much should each plant ship to each customer
in order to minimize the total cost? Linear programming can provide the answer. This type
of linear programming problem is called a transportation problem.

This kind of application normally needs two kinds of functional constraints. One kind
specifies that the amount of the product produced at each plant must equal the total
amount shipped to customers. The other kind specifies that the total amount received
from the plants by each customer must equal the amount ordered. These are fixed-
requirement constraints, which makes the problem a fixed-requirements problem.
However, there also are variations of this problem where resource constraints or benefit
constraints are needed.
Formulation of the Problem in Linear Programming Terms
We need to identify the activities and requirements of this transportation problem to
formulate it as a linear programming problem.
V/ Assignment problems:
We now turn to another special type of linear programming problem called assignment
problems. As the name suggests, this kind of problem involves making assignments.
Frequently, these are assignments of people to jobs. Thus, many applications of the
assignment problem involve aiding managers in matching up their personnel with tasks
to be performed. Other applications might instead involve assigning machines, vehicles,
or plants to tasks.

The model for Assignement Problems

Given a set of tasks to be performed and a set of assignees who are available to perform
these tasks, the problem is to determine which assignee should be assigned to each task.
To fit the model for an assignment problem, the following assumptions need to be satisfied:
1. The number of assignees and the number of tasks are the same.
2. Each assignee is to be assigned to exactly one task.
3. Each task is to be performed by exactly one assignee.
4. There is a cost associated with each combination of an assignee
performing a task.
5. The objective is to determine how all the assignments should be made to
minimize the total cost.

CHAPTER 5
Sensitivity report:
I/ Changes in Objective Function Coefficient:
1. What happens when one objective function coefficient changes (increase/
decrease)?

2. What happens when objective function coefficients (2 coefficients or more) changes

(increase/ decrease) simultaneously?
The 100 Percent Rule for Simultaneous Changes in Objective Function Coefficients:
If simultaneous changes are made in the coefficients of the objective function, calculate
for each change the percentage of the allowable change (increase or decrease) for
that coefficient to remain within its allowable range. If the sum of the percentage changes
does not exceed 100 percent, the original optimal solution definitely will still be
optimal. (If the sum does exceed 100 percent, then we cannot be sure)
ð CONCLUSION: THE ORIGINAL OPTIMAL SOLUTION DEFINITELY WILL STILL
BE OPTIMAL OR WE ARE NOT SURE.

II/ Changes in a Constraint (RHS):

1. What happens when one RHS of a constraint changes (increase/ decrease)?
Shadow price: Given an optimal solution and the corresponding value of the objective
function for a linear programming model, the shadow price for a functional constraint is
the rate at which the value of the objective function changes by 1 unit change on
the right-hand-side.
2. What happens when RHSs of some constraints (2 RHSs or more) changes
(increase/ decrease) simultaneously?

The 100 Percent Rule for Simultaneous Changes in Right-Hand Sides:

The shadow prices remain valid for predicting the effect of simultaneously
changing the right-hand sides of some of the functional constraints as long as
the changes are not too large. To check whether the changes are small enough,
calculate for each change the percentage of the allowable change (decrease or
increase) for that right-hand side to remain within its allowable range. If the sum of the
percentage changes does not exceed 100 percent, the shadow prices definitely will
still be valid. (If the sum does exceed 100 percent, then we cannot be sure.)

ð CONCLUSION: THE SHADOW PRICE(S) DEFINITELY WILL BE VALID OR WE

ARE NOT SURE.

CHAPTER 6
I/ Minimum-cost flow problems:
Terminology
1. The model for any minimum-cost flow problem is represented by a network with flow
passing through it.
2. The circles in the network are called nodes.
3. Each node where the net amount of flow generated (outflow minus inflow) is a fixed -
positive number is a supply node.

4. Each node where the net amount of flow generated is a fixed negative number is a
demand node.
5. Any node where the net amount of flow generated is fixed at zero is a transshipment
node.
6. The arrows in the network are called arcs.
7. The maximum amount of flow allowed through an arc is referred to as the capacity of
that arc.
Using this terminology, the general characteristics of minimum-cost flow problems (the
model for this type of problem) can be described in terms of the following assumptions.
Assumptions of a Minimum-Cost Flow Problem
1. At least one of the nodes is a supply node, so each one specifies its own fixed
positive - number for the net amount of flow generated there.
2. At least one of the other nodes is a demand node, so each one specifies its own
fixed - negative number for the net amount of flow generated there.
3. All the remaining nodes are transshipment nodes, so each one specifies a fixed
value of zero for the net amount of flow generated there.
4. Flow through an arc is only allowed in the direction indicated by the arrowhead,
where the maximum amount of flow is given by the capacity of that arc. (If flow can occur
in both directions, this would be represented by a pair of arcs pointing in opposite
directions.)

5. The network has enough arcs with sufficient capacity to enable all the flow generated
at the supply nodes to reach all the demand nodes.
6. The cost of the flow through each arc is proportional to the amount of that flow, where
the cost per unit flow is known.
7. The objective is to minimize the total cost of sending the available supply through
the network to satisfy the given demand. (An alternative objective is to maximize the total
profit from doing this.)
A solution for this kind of problem needs to specify how much flow is going through each
arc. To be a feasible solution, the amount of flow through each arc cannot exceed the
capacity of that arc and the net amount of flow generated at each node must equal the
specified amount for that node. The following property indicates when the problem will
have feasible solutions.

Feasible Solutions Property: Under the above assumptions, a minimum-cost flow problem
will have feasible solutions if and only if the sum of the supplies from its supply nodes
equals the sum of the demands at its demand nodes.
For many applications of minimum-cost flow problems, management desires a solution
with integer values for all the flow quantities (e.g., integer numbers of full truckloads
along each shipping lane). The model does not include any constraints that require this
for feasible solutions. Fortunately, such constraints are not needed because of the
following property.
Integer Solutions Property: As long as all its supplies, demands, and arc capacities have
integer values, any minimum-cost flow problem with feasible solutions is guaranteed to
have an optimal solution with integer values for all its flow quantities.
Some Applications

II/ Maximum flow problem:

Assumptions of a Maximum Flow Problem

1. All flow through the network originates at one node, called the source, and
terminates at one other node, called the sink.
2. All the remaining nodes are transshipment nodes.
3. Flow through an arc is only allowed in the direction indicated by the arrowhead,
where the maximum amount of flow is given by the capacity of that arc. At the source, all
arcs point away from the node. At the sink, all arcs point into the node.
4. The objective is to maximize the total amount of flow from the source to the sink.
This amount is measured in either of two equivalent ways, namely, either the amount
leaving the source or the amount entering the sink.
The source and sink of a maximum flow problem are analogous to the supply nodes and
demand nodes of a minimum-cost flow problem. These are the only nodes in both
problems that do not have conservation of flow (flow out equals flow in). Like the supply
nodes, the source generates flow. Like the demand nodes, the sink absorbs flow.
However, there are two differences between these nodes in a minimum-cost flow problem
and the corresponding nodes in a maximum flow problem.
One difference is that, whereas supply nodes have fixed supplies and demand nodes
have fixed demands, the source and sink do not. The reason is that the objective is to
maximize the flow leaving the source and entering the sink rather than fixing this amount.
The second difference is that, whereas the number of supply nodes and the number of
demand nodes in a minimum-cost flow problem may be more than one, there can be only
one source and only one sink in a maximum flow problem.
Some Applications
The applications of maximum flow problems and their variants are somewhat similar to
those for minimum-cost flow problems described in the preceding section when
management’s objective is to maximize flow rather than to minimize cost. Here are some
typical kinds of applications.
1. Maximize the flow through a distribution network.
2. Maximize the flow through a company’s supply network from its vendors to its
processing facilities.
3. Maximize the flow of oil through a system of pipelines.
4. Maximize the flow of water through a system of aqueducts.
5. Maximize the flow of vehicles through a transportation network.

III/ The shortest path:

Assumptions of a Shortest Path Problem
1. You need to choose a path through the network that starts at a certain node, called
the origin, and ends at another certain node, called the destination.
2. The lines connecting certain pairs of nodes commonly are links (which allow travel
in either direction), although arcs (which only permit travel in one direction) also are
allowed.
3. Associated with each link (or arc) is a nonnegative number called its length. (Be
aware that the drawing of each link in the network typically makes no effort to show its
true length other than giving the correct number next to the link.)
4. The objective is to find the shortest path (the path with the minimum total length)
from the origin to the destination.

Some Applications
Not all applications of shortest path problems involve minimizing the distance traveled
from the origin to the destination. In fact, they might not even involve travel at all. The
links (or arcs) might instead represent activities of some other kind, so choosing a path
through the network corresponds to selecting the best sequence of activities. The
numbers giving the “lengths” of the links might then be, for example, the costs of the
activities, in which case the objective would be to determine which sequence of activities
minimizes the total cost. Here are three categories of applications.
1. Minimize the total distance traveled.
2. Minimize the total cost of a sequence of activities.
3. Minimize the total time of a sequence of activities.

CHAPTER 9
Decision Analysis Terminology
Decision analysis has a few special terms.
The decision maker is the individual or group responsible for making the decision (or
sequence of decisions) under consideration.
The alternatives are the options for the decision to be made by the decision maker.

The outcome of the decision to be made will be affected by random factors that are
outside the control of the decision maker. These random factors determine the situation
that will be found when the decision is executed. Each of these possible situations is
referred to as a possible state of nature.
The decision maker generally will have some information about the relative likelihood of
the possible states of nature. This information may be in the form of just subjective
estimates based on the experience or intuition of an individual, or there may be some
degree of hard evidence involved. When these estimates are expressed in the form of
probabilities, they are referred to as the prior - probabilities of the respective states of
nature. Although these are unlikely to be the true probabilities based on more information,
they are the best available estimates of the probabilities prior to obtaining more
information.
Each combination of a decision alternative and a state of nature results in some outcome.
The payoff is a quantitative measure of the value to the decision maker of the
consequences of the outcome. In most cases, the payoff is expressed as a monetary
value, such as the profit.
The Payoff Table
When formulating the problem, it is important to identify all the relevant decision
alternatives and the possible states of nature. After identifying the appropriate measure
for the payoff from the perspective of the decision maker, the next step is to estimate the
payoff for each combination of a decision alternative and a state of nature. These payoffs
then are displayed in a payoff table.
I/ Decision criteria:
Decision Making without Probabilities: The Maximax Criterion
The maximax criterion is the decision criterion for the eternal optimist. It says to focus
only on the best that can happen to us. Here is how this criterion works:
1. Identify the maximum payoff from any state of nature for each decision alternative.
2. Find the maximum of these maximum payoffs and choose the corresponding
decision alternative.
=> The maximax criterion always chooses the decision alternative that can give the
largest possible payoff.
Decision Making without Probabilities: The Maximin Criterion
The maximin criterion is the criterion for the total pessimist. In contrast to the maximax
criterion, it says to focus only on the worst that can happen to us. Here is how this criterion
works:
1. Identify the minimum payoff from any state of nature for each decision alternative.
2. Find the maximum of these minimum payoffs and choose the corresponding
decision alternative.
=> The maximin criterion always chooses the decision alternative that provides the best
guarantee for its minimum possible payoff.
Decision Making with Probabilities: The Maximum Likelihood Criterion
The maximum likelihood criterion says to focus on the most likely state of nature as
follows.
1. Identify the state of nature with the largest prior probability.
2. Choose the decision alternative that has the largest payoff for this state of nature.
=> The maximum likelihood criterion assumes that the most likely state of nature will
occur and chooses accordingly. This criterion ignores all the payoffs except for the most
likely state of nature.
Decision Making with Probabilities: Bayes’ Decision Rule
Bayes’ decision rule directly uses the prior probabilities of the possible states of nature
as summarized below.
1. For each decision alternative, calculate the weighted average of its payoffs by
multiplying each payoff by the prior probability of the corresponding state of nature and
then summing these products. Using statistical terminology, refer to this weighted average
as the expected payoff (EP) for this decision alternative.
2. Bayes’ decision rule says to choose the alternative with the largest expected payoff.
=> On the average, Bayes’ decision rule provides larger payoffs in the long run than any
other criterion. More recently, it has become somewhat popular to also call this criterion
the expected monetary value (EMV) criterion.
II/ Decision tree:
In the terminology of decision trees, the junction points are called nodes (or forks) and
the lines emanating from the nodes are referred to as branches. A distinction is then
made between the following two types of nodes.
A decision node, represented by a square, indicates that a decision needs to be made
at that point in the process.

An event node (or chance node), represented by a circle, indicates that a random event
occurs at that point.
III/ Checking whether to obtain more information:
Prior probabilities may provide somewhat inaccurate estimates of the true probabilities
of the states of nature. The quickest way to check this is to pretend that it is possible for
the same amount of money to actually determine which state is the true state of nature
(“perfect information”) and then determine whether obtaining this information would make
this expenditure worthwhile. If having perfect information would not be worthwhile, then it
definitely would not be worthwhile to spend this money just to learn more about the
probabilities of the states of nature.
The key quantities for performing this analysis are
EP (without more info) = Expected payoff from applying Bayes’ decision rule with the
original prior probabilities

EP (with perfect info) = Expected payoff if the decision could be made after learning the
true state of nature
EVPI = Expected value of perfect information
C = Cost of obtaining more information
The expected value of perfect information is calculated as
EVPI = EP (with perfect info) − EP (without more info)
After calculating EP (with perfect info) and then EVPI, the last step is to compare EVPI
with C.
If C > EVPI, then it is not worthwhile to obtain more information.
If C ≤ EVPI, then it might be worthwhile to obtain more information.
IV/ Using new information to update the probabilities:
The prior probabilities of the possible states of nature often are quite subjective in nature,
so they may be only very rough estimates of the true probabilities. Fortunately, it
frequently is possible to do some additional testing or surveying (at some expense) to
improve these estimates. These improved estimates are called posterior probabilities.
To use either finding to calculate the posterior probability, it is necessary to estimate the
probability of obtaining this finding for each state of nature. The notation used in the table
for each of these estimated probabilities is

P (finding | state) = Probability that the indicated finding will occur, given that the
state of nature is the indicated one
This kind of probability is referred to as a conditional probability, because it is
conditioned on being given the state of nature.
The next step is to use these probabilities and the probabilities in Table 9.7 to obtain a
combined probability called a joint probability. Each combination of a state of nature
and a finding from the seismic survey will have a joint probability that is determined by
the following formula.
P (state and finding) = P (state) * P (finding | state)
The next step is to use these probabilities to find each probability of just a particular finding
without specifying the state of nature. Since any finding can be obtained with any state of
nature, the formula for calculating the probability of just a particular finding is
P(finding) = P(Oil and finding) + P(Dry and finding)
Finally, we now are ready to calculate each posterior probability of a particular state of
nature.
P(state | finding) = P(state and finding) / P(finding)
V/ Expected value of sample information (EVSI):
Suppose that there is uncertainty about the cost to obtain more information. How
would this change the analysis?
This uncertainty would raise the new key question of how much value would be added by
having more information. In other words, how large can the cost to obtain more
information be before it would no longer be worthwhile to perform it. This quantity is
referred to as the expected value of sample information (or EVSI for short), where “sample
information” in this case refers to the extra information.
The analysis needed to determine EVSI would begin by identifying two key quantities,
EP (with more info) = Expected payoff (excluding the cost of the survey) when the
survey is done
EP (without more info) = Expected payoff when the survey is not done
where Bayes’ decision rule is applied to find both quantities.
Then we can calculate the expected value of sample information (EVSI) as
EVSI = EP (with more info) − EP (without more info)
Let
C = Best available estimate of the cost to obtain more information.

The final step in the analysis is to compare C and EVSI.

If C < EVSI, then perform the test/ survey to obtain more information.
If C ≥ EVSI, then do not perform the test/ survey to obtain more information.

Eg. Goferbroke
EP (with more info) can be obtained from the decision tree by referring to the expected
payoff of 123 except that the (unknown) cost of the seismic survey should not be included,
so 30 (the originally assumed cost of the survey) should be added to this expected payoff.
Therefore,
EP (with more info) = 123 + 30 = 153
EP (without more info) is described as the expected payoff from applying Bayes’
decision rule with the original prior probabilities (so not using the seismic survey), where
this expected payoff was found to be 100. So
EP (without more info) = 100
Now we can calculate the expected value of sample information (EVSI) as
EVSI = EP (with more info) − EP (without more info)
= 153 − 100
= 53
Let
C = Best available estimate of the cost of the seismic survey (in thousands of dollars)
The final step in the analysis is to compare C and EVSI.
If C < EVSI, then perform the seismic survey.
If C ≥ EVSI, then do not perform the seismic survey.

CHAPTER 10 – (đọc thêm cách dùng Excel để dự báo với nhiều phương pháp dự
báo khác nhau – trung bình động, trung bình động có trọng số, san bằng hàm mũ…)

CHÚC CÁC BẠN LÀM BÀI THI TỐT!