Professional Documents
Culture Documents
Decision Tree
● A decision tree is a chronological
representation of the decision
problem. Choose the decision alternative with
● Each decision tree has two types of the largest EV. Build the large complex.
nodes: round nodes correspond to
the states of nature while square Expected Value of Perfect Information
nodes correspond to the decision ● Frequently information is available
alternatives. which can improve the probability
● The branches leaving each round estimates for the states of nature.
node represent the different states of ● The expected value of perfect
nature while the branches leaving information is the increase in the
expected profit that would result if Sensitivity Analysis
one knew with certainty which state ● Sensitivity analysis can be used to
of nature would occur. determine how changes to the
● The EVPI provides an upper bound following inputs affect the
on the expected value of any sample recommended decision alternative:
or survey information. ● probabilities for the states of
nature
● values of the payoffs
● If a small change in the value of one
of the inputs causes a change in the
recommended decision alternative,
extra effort and care should be taken
in estimating the input value.
Influence Diagram
Sample Information
PDC has developed the following branch
probabilities.
Decision Strategy
❖ A decision strategy is a sequence of
If the market research report is decisions and chance outcomes
unfavorable: where the decisions chosen depend
on the yet-to-be-determined
outcomes of chance events.
❖ The approach used to determine the
optimal decision strategy is based
on a backward pass through the
decision tree using the following PDC’s Decision Strategy
steps: PDC’s optimal decision strategy is:
● At chance nodes, compute ● Conduct the market research study.
the expected value by ● If the market research report is
multiplying the payoff at the favorable, construct the large
end of each branch by the condominium complex.
corresponding branch ● If the market research report is
probabilities. unfavorable, construct the medium
● At decision nodes, select the condominium complex.
decision branch that leads to
the best expected value. This Risk Profile
expected value becomes the
expected value at the
decision node.
Decision Tree 2
Posterior Probabilities
Scatter Plots
A scatter plot is designed to show a
relationship between two variables by
graphing a series of observations on a
two-dimensional graph- one variable on the
X-axis, the other on the Y-axis. average of the cross- products is computed
by dividing the sum by n 1, or 4. The use of
Scatter Plot n - 1 for covariance is done by statisticians
to ensure an unbiased estimate. Interpreting
a covariance number is difficult for those
who are not statistical experts. The 99.64
we computed for this example has a sign of
"returns squared" since the numbers were
percentage returns, and a return squared is
not an intuitive concept. The fact that
Cov(X,Y) of 99.64 was greater than 0 does
indicate a positive or linear relationship
between X and Y. Had the covariance been
a negative number, it would imply an
Sample Covariance inverse relationship, while 0 means no
To quantify a linear relationship relationship. Thus 99.64 indicates that the
between two variables, we start by finding returns have positive co-movement (when
the covariance of a sample of paired one moves higher so does the other), but
observations. A sample covariance between doesn't offer any information on the extent
two random variables X and Y is the of the co-movement.
average value of the cross-product of all
observed deviations from each respective Sample Correlation Coefficient
sample mean. A cross- product, for the ith By calculating a correlation coefficient, we
observation in a sample, is found by this essentially convert a raw covariance
calculation: (ith observation of X - sample number into a standard format that can be
mean of X)" (ith observation of Y-sample more easily interpreted to determine the
mean of Y). The covariance is the sum of all extent of the relationship between two
cross-products, divided by (n-1). variables. The formula for calculating a
sample correlation coefficient (r) between
To illustrate, take a sample of live paired two random variables
observations of annual returns for two
mutual funds, which we will label X and Y
X and Y is the following:
r-(covariance between X, Y)/(sample
standard deviation of X)* (sample std. dev.
of Y).
Answer:
Interpreting Covariance
Linear Correlation
Correlation coefficient
■ Pearson's Correlation Coefficient is
standardized covariance (unitless):
Correlation
● Measures the relative strength of the
linear relationship between two
variables
● Unit-less
● Ranges between -1 and 1
● The closer to -1, the stronger the
negative linear relationship
● The closer to 1, the stronger the
positive linear relationship
● The closer to 0, the weaker any
positive linear relationship
Calculating by hand… Regression Analysis
● A set of statistical methods used to
estimate the relationships between a
dependent variable and one or more
independent variables.
● It can be used to assess the strength
of the relationship between variables
and for modeling the future
relationship between them.
Simpler calculation formula…
Variations of Regression Analysis:
● Simple Linear Regression
● Multiple Linear Regression
● Nonlinear Regression
Y-Dependent variable
X-Independent (explanatory) variable
a - Intercept
b-Slope
e-Residual (error)
The sample correlation coefficient follows a
T-distribution with n- 2 degrees of freedom Fundamental Assumptions of Multiple
(since you have to estimate the standard Linear Regression Analysis
error). ● The dependent and independent
variables show a linear relationship
*note, like a proportion, the variance of the between the slope and the intercept.
correlation coefficient depends on the ● The independent variable is not
correlation coefficient itself substitute in random.
estimated r ● The value of the residual (error) is
zero.
● The value of the residual (error) is
constant across all observations.
● The value of the residual (error) is ● Forecasting Revenues and
not correlated across all Expenses
observations.
● The residual (error) values follow the
normal distribution.
● Non-collinearity: Independent
variables should show a minimum of
correlation with each other. If the
independent variables are highly
correlated with each other, it will be
difficult to assess the true
relationships between the dependent
and independent variables.
Nonlinear Regression
● Nonlinear regression analysis is
commonly used for more
complicated data sets in which the
dependent and independent
variables show a nonlinear
relationship Chapter 5. Learning Curve Analysis
Chapter 5
Examples of Applications of Regression
Analysis in Finance
● Beta and CAPM.
Chapter 6. Linear Programming Guidelines for Model Formulation
Chapter 2 Problem formulation or modeling is the
2.1 - A Simple Maximization Problem process of translating a verbal statement of
2.2 - Graphical Solution Procedure a problem into a mathematical statement.
2.3 - Extreme Points and the Optimal ● Understand the problem thoroughly.
Solution ● Describe the objective.
2.4 - Computer Solution of the Par, Inc., ● Describe each constraint.
Problem ● Define the decision variables.
2.5 - A Simple Minimization Problem ● Write the objective in terms of the
2.6 - Special Cases decision variables.
2.7 - General Linear Programming Notation ● Write the constraints in terms of the
decision variables.
Linear Programming
● Linear programming has nothing to A Simple Maximization Problem
do with computer programming. Par, Inc. is a small manufacturer of golf
● The use of the word “programming” equipment and supplies whose
here means “choosing a course of management has decided to move into the
action.” market for medium- and high-priced golf
● Linear programming involves bags. Par, Inc.’s distributor has agreed to
choosing a course of action when buy all the golf bags Par, Inc., produces
the mathematical model of the over the next three months.
problem contains only linear Each golf bag produces will require the
functions. following operations:
● The maximization or minimization of 1. Cutting and dyeing the material
some quantity is the objective in all 2. Sewing
linear programming problems. 3. Finishing (inserting umbrella holder,
● All LP problems have constraints club separators, etc.)
that limit the degree to which the 4. Inspection and packaging
objective can be pursued.
● A feasible solution satisfies all the This production information is summarized
problem’s constraints. in this table:
● An optimal solution is a feasible
solution that results in the largest
possible objective function value
when maximizing (or smallest when
minimizing).
● A graphical solution method can be
used to solve a linear program with
two variables. ● Par, Inc.’s production is constrained
by a limited number of hours
available in each department. The
director of manufacturing estimates
that 630 hours for cutting and
dyeing, 600 hours for ewing, 708
hours for finishing, and 135 hours for We continue by identifying the solution
inspection and packaging will be points satisfying each of the other three
available for the production of golf constraints.
bags during the next three months.
● The accounting department
analyzed the production data and
arrived at prices for both bags that
will result in a profit contribution 1 of
$10 for every standard bag and $9
for every deluxe bag produced.
The complete model for the Par, Inc., The graph shown identifies the feasible
problem is as follows: region:
Computer Solution
Feasible Region Furthermore, any point on the line
● The feasible region for a connecting the two optimal extreme points
two-variable LP problem can be also provides an optimal solution.
nonexistent, a single point, a line, a
polygon, or an unbounded area. For example, the solution point (S= 420,
● Any linear program falls in one of D=336), which is halfway between the two
four categories: extreme points, also provides the optimal
● is infeasible objective function value of 6.3S+ 9D =
● has a unique optimal solution 6.3(420) + 9(336) = 5670.
● has alternative optimal
solutions A linear programming problem with
● has an objective function that alternative optimal solutions is generally a
can be increased without good situation for the manager or decision
bound maker. It means that several combinations
● A feasible region may be unbounded of the decision variables are optimal and
and yet there may be optimal that the manager can select the most
solutions. This is common in desirable optimal solution.
minimization problems and is
possible in maximization problems. Infeasibility
● No solution to the LP problem
Special Cases satisfies all the constraints, including
Alternative Optimal Solutions the non-negativity conditions.
In the graphical method, if the ● Graphically, this means a feasible
objective function line is parallel to a region does not exist.
boundary constraint in the direction of ● Causes include:
optimization, there are alternate optimal ● A formulation error has been
solutions, with all points on this line made.
segment being optimal. ● Management's expectations
are too high.
Let's return to the Par, Inc., problem. ● Too many restrictions have
However, now assume that the profit for the been placed on the problem
standard golf bag (S) has decreased to (i.e. the problem is
$6.30. The revised objective function over-constrained).
becomes 6.3S+9D.
The objective function values at these two Unbounded
extreme points are identical: ● The solution to a maximization LP
6.3S+9D = 6.3(300)+9(420)=5670 problem is unbounded if the value of
and the solution may be made
6.3S+9D = 6.3(540)+9(252)=5670 indefinitely large without violating
any of the constraints.
● For real problems, this is the result
of improper formulation. (Quite likely,
a constraint has been inadvertently
omitted.)
General Linear Programming Notation Chapter 4
We selected decision-variable names of S 4.1 - Marketing Applications
and D in the Par, Inc., problem and A and B 4.2 - Financial Applications
in the M&D Chemicals problem to make it 4.3 - Operations Management Applications
easier to recall what these decisions were.
variables represented in the problem. Marketing Applications
Media Selection
Although this approach works well for linear ● One application of linear
programs involving a small number of programming in marketing is media
decision variables, it can become difficult selection.
when dealing with problems involving a ● LP can be used to help marketing
large number of decision variables. managers allocate a fixed budget to
various advertising media.
A more general notation that is often used ● The objective is to maximize reach,
for linear programs uses the letter x with a frequency, and quality of exposure.
subscript. ● Restrictions on the allowable
allocation usually arise during
In the Par, Inc., problem, we could have consideration of company policy,
defined the decision variables: contract requirements, and media
availability.
X = number of standard bags
X2 = number of deluxe bags Relax-and-Enjoy Lake Development
Corporation is developing a lakeside
In the M&D Chemicals problem, the same community at a privately owned lake.
variable names would be used, but their
definitions would change: The primary market for the lakeside lots and
X1 = number of gallons of product A homes includes all middle- and
X2 = number of gallons of product B upper-income families within approximately
100 miles of the development.
A disadvantage of using general notation for
decision variables is that we are no longer Relax-and-Enjoy employed the advertising
able to easily identify what the decision firm of Boone, Phillips, and Jackson (BP&J)
variables actually represent in the to design the promotional campaign.
mathematical model.
BP&J collected data on the number of
The advantage of general notation is that potential customers reached, the cost per
formulating a mathematical model for a advertisement, the maximum number of
problem that involves a large number of times each medium is available, and the
decision variables is much easier. exposure quality rating for each of the five
media.
The decision to be made is how many times The optimal solution to this five-variable,
to use each medium. We begin by defining nice-constraint linear programming model is
the decision variables: shown here.
Let:
Basic Solution
To determine a basic solution, set n - m of
the variables equal to zero, and solve the m
linear constraint equations for the remaining
Tableau Form
m variables.
A basic feasible solution to the system of m
We shall refer to the n 2 m variables set
linear constraint equations and n variables
equal to zero as the nonbasic variables and
is required as a starting point for the simplex
the remaining m variables as the basic
method. The purpose of tableau form is to
variables. Thus, in the preceding example,
provide an initial basic feasible solution.
x2 and s1 are the nonbasic variables, and x1,
Recall that for the HighTech problem, the
s2, and s3 are the basic variables.
standard-form representation is
When these conditions are satisfied, exactly Setting Up The Initial Simplex Tableau
one basic variable with a coefficient of 1 is After a linear programming problem has
associated with each constraint equation, been converted to tableau form, we have an
and for each of the m constraint equations, initial basic feasible solution that can be
it is a different basic variable. Thus, if the used to begin the simplex method.
n-m nonbasic variables are set equal to
zero, the values of the basic variables are To provide a convenient means for
the values of the right-hand sides of the performing the calculations required by the
constraint equations. simplex method, we will first develop what is
The second property that enables us to find referred to as the initial simplex tableau.
a basic feasible solution requires the values
of the right-hand sides of the constraint Part of the initial simplex tableau is a table
equations be nonnegative. This containing all the coefficients shown in the
nonnegativity ensures that the basic tableau form of a linear program. If we
solution obtained by setting the basic adopt the general notation
variables equal to the values of the cj =objective function coefficient for variable j
righthand sides will be feasible. bj=right-hand-side value for constraint j
aij=coefficient associated with variable j in
If a linear programming problem satisfies constraint / we can show this portion of the
these two properties, it is said to be in initial simplex tableau as follows:
tableau form.
To help us recall that each of the columns To move from an initial basic feasible
contains the coefficients for one of the solution to a better basic feasible solution,
variables, we write the variable associated the simplex method must generate a new
with each column directly above the column. basic feasible solution that yields a better
By adding the variables we obtain value for the objective function.
To find out whether it is possible, we add Recall that the values in the cB column of
two rows to the bottom of the tableau. The the simplex tableau are the objective
first row, labeled zj represents the decrease function coefficients for the current basic
in the value of the objective function that will variables.
result if one unit of the variable
corresponding to the jth column of the A Hence, to compute the values in the zj row
matrix is brought into the basis. The second (the decrease in value of the objective
row, labeled cj - zj represents the net function when xj is increased by one), we
change in the value of the objective function form the sum of the products obtained by
if one unit of the variable corresponding to multiplying the elements in the cB column by
the jth column of the A matrix is brought into the corresponding elements in the jth column
the solution. We refer to the cj - zj row as the of the A matrix. Doing these calculations we
net evaluation row. obtain
Transportation Problem
● The transportation problem arises
frequently in planning for the
Management would like to determine X13 = the number of units shipped
how much of its production should be from origin 1 (Cleveland) to destination 3
shipped from each plant to each distribution (St. Louis)
center. X14 = the number of units shipped
from origin 1 (Cleveland) to destination 4
Here is a graph showing the 12 (Lexington)
distribution routes Foster can use. X21 = the number of units shipped
● Such a graph is called a network. from origin 2 (Bedford) to destination 1
● The circles are referred to as nodes (Boston)
and the lines connecting the nodes
as arcs.
● Note that the direction of flow (from
origin to destination) is indicated by
the arrows.
● Each origin and destination is
represented by a node, and each
possible shipping route is
represented by an arc.
● The amount of the supply is written
next to each origin node, and the
amount of the demand is written
next to each destination node.
● The goods shipped from the origins
to the destinations represent the flow
in the network.
The sum of the completion times for Because the assignment problem
the 3 project leaders provide the total days can be viewed as a special case of the
required to complete the three assignments. transportation problem, the problem
Thus, the objective function is variations that may arise in an assignment
MIN 10x11 + 15x12 + 9x13 + 18x22 +
5x23 + 6x31 + 14x32 + 3x33
problem parallel those for the transportation could happen if an agent did not have the
problem. experience necessary for one or more of the
Specifically, we can handle tasks.
1. Total number of agents (supply) not
equal to the total number of tasks Shortest-Route Problem
(demand) The shortest-route problem is
2. A maximization objective function concerned with finding the shortest path in a
3. Unacceptable assignments network from one node (or set of nodes) to
another node (or set of nodes).
If the total number of agents (supply) ● If all arcs in the network have
are not equal to the total number of tasks nonnegativity values then a labeling
(demand): algorithm can be used to find the
● This is analogous to total supply not shortest paths from a particular node
equaling total demand in a to all other nodes in the network.
transportation problem. ● The criterion to be minimized in the
● The extra agents simply remain shortest-route problem is not limited
unassigned in the linear to distance even though the term
programming solution. “shortest” is used in describing the
● If the number of tasks exceeds the procedure. Other criteria include
number of agents, the linear time and cost. (Neither time nor cost
programming model will not have a are necessarily linearly related to
feasible solution. By adding two distance.)
dummy project leaders, we can
create a new assignment problem Maximal Flow Problem
with the number of project leaders The objective in a maximal flow
equal to the number of clients. The problem is to determine the maximum
objective function coefficients for the amount of flow (vehicles, messages, fluid,
assignment of dummy project etc.) that can enter and exit a network
leaders would be zero so that the system in a given period of time.
value of the optimal solution would ● In this problem, we attempt to
represent the total number of days transmit flow through all arcs of the
required by the assignments actually network as efficiently as possible.
made. ● The amount of flow is limited due to
capacity restrictions on the various
If the assignment alternatives are arcs of the network. For example,
evaluated in terms of revenue or profit highway types limit vehicle flow in a
rather than time or cost, the linear transportation system, while pipe
programming formulation can be solved as sizes limit oil flow in an oil
a maximization rather than a minimization distribution system.
problem. ● The maximum or upper limit on the
If one or more assignments are flow in an arc is referred to as the
unacceptable, the corresponding decision flow capacity of the arc.
variable can be removed from the linear ● Even though we do not specify
programming formulation. This situation capacities for the nodes, we do
assume that the flow out of a node is
equal to the flow into the node.