Management Scientist User's Guide

THE MANAGEMENT SCIENTIST
VERSION 5.0
FOR WINDOWS® 95 AND WINDOWS® 98
David R. Anderson
University of Cincinnati
Dennis J. Sweeney
University of Cincinnati
Thomas A. Williams
Rochester Institute of Technology
COPYRIGHT ©2003 by South-Western, a division of Thomson Learning.

The Thomson Learning logo is a registered trademark used herein under license.
i
TABLE OF CONTENTS
Preface
Chapter 1 Introduction
Chapter 2 Linear Programming
Chapter 3 Transportation Problem
Chapter 4 Assignment Problem
Chapter 5 Integer Linear Programming
Chapter 6 Shortest Route
Chapter 7 Minimal Spanning Tree
Chapter 8 PERT/CPM
Chapter 9 Inventory
Chapter 10 Waiting Lines
Chapter 11 Decision Analysis
Chapter 12 Forecasting
Chapter 13 Markov Processes
ii
PREFACE
This user’s manual accompanies version 5.0 of The Management Scientist, a personal
computer software package developed by David R. Anderson, Dennis J. Sweeney, and
Thomas A. Williams. The software, developed to assist students enrolled in quantitative
methods and management science courses, can be used to solve a wide variety of textbook
problems as well as small-scale problems encountered in practice. Use of The Management
Scientist demonstrates the valuable role of the computer in applying quantitative methods to
decision problems.
The new version 5.0, available for Windows 95 and Windows 98 environments, has a
significantly improved user interface as well as more robust solution algorithms.
We take this opportunity to acknowledge the major contribution made by Amy Anneken
in the development of Versions 4.0 and 5.0. Also, we are indebted to Daniel Joseph and
Bradley Beck who assisted with previous versions of the software. At this time we do not
know of errors but, of course, take responsibility for any that remain.
If you have any comments and suggestions regarding The Management Scientist
software, we welcome your input. Please send comments and suggestions to
Dr. David R. Anderson

David.anderson.uc.edu
David R. Anderson
Dennis J. Sweeney
Thomas A. Williams
iii
CHAPTER 1
INTRODUCTION
Welcome to version 5.0 of The Management Scientist. You are ready to

install and begin using a PC software package designed to solve the
following quantitative models for managerial decision making:
Linear programming
Transportation
Assignment
Integer linear programming
Shortest route
Minimal spanning tree
PERT/CPM
Inventory
Waiting lines
Decision analysis
Forecasting
Markov processes
You do not have to be an expert in quantitative analysis to use this

soft ware package. If you have a general understanding of the above
types of problems, you simply select the program module that is
appropriate for the problem you wish to solve. The corresponding chapter
in this manual will guide you through an example problem and familiarize
you with the data input, special features, and output information provided
by the module. After following the step-by-step procedure for the example
problem, you should be able to input, solve, and interpret the results for
other problems that use that module.
1.1 USING THIS MANUAL

Most of us, after obtaining a new software package, would like to begin
solving problems immediately; we do not want to read the complete
manual just to get ready to use the software. The manual for The
Management Scientist was prepared with this thought in mind. We
recommend taking the following steps to minimize the start-up effort and
thus be able to begin using the software as soon as possible.
1
1.Follow the directions in Section 1.2 to install The Management
Scientist on a personal computer using Windows 95, Windows 98, or
Windows NT.
2.Continue to read, or at least skim, the rest of Chapter 1 to obtain an

overview of the entire software package and most of its features.
3.Select any one of the 12 modules of The Management Scientist to

try. The chapters of the manual are independent of each other, so
you should have no difficulty going directly to Chapter 9 if, for
example, you want to use the inventory module first or going
directly to Chapter 4 if you would like to try the assignment module.
The discussion describing the module introduces you to the type of
problems that can be solved and then shows you how to create and
solve an example problem.
1.2 INSTALLATION
Version 5.0 of The Management Scientist will run on personal computers
running Windows 95, Windows 98, or Windows NT. Installing the software
on your computer is easy. Install The Management Scientist as follows:
1.Place the disk in the CD-ROM drive.
2.In Windows from the Start taskbar, choose Run.
3.On the Command line, type D:\ Setup. If your CD-ROM drive is not
D:, substitute the appropriate letter.
4.Click the OK button or press the Enter key.
5.Follow the prompts of the installation.
Running The Management Scientist

Once The Management Scientist has been installed it is ready to be used.
To run MS 5.0 after it has been installed, go to Programs from the Start
button. Go to The Management Scientist group and click on The
Management Scientist, or its logo. The opening screen will then appear.
Just click Continue to begin working with one of the 12 modules.
1.3 OVERVIEW OF THE SOFTWARE PACKAGE

The Management Scientist software package consists of 12 computer pro
grams, called modules, that use quantitative methods to develop
2
decision-making information. A brief description of each of the modules
including problem-size information is provided below. More detailed
information about each of the modules is presented in the remaining
chapters.
Modules Available in The Management Scientist

Linear Programming. This program will maximize or minimize the value
of a linear objective function subject to a set of linear constraints. The
program can handle up to 100 variables and up to 50 constraints.
Sensitivity analysis information is provided.
Transportation. This program will find the minimum cost or maximum

profit schedule for the transportation of units from several supply origins
to several demand destinations. Transportation problems with up to 30
origins and up to 30 destinations are permitted.
Assignment. This program will minimize the cost or maximize the rev-
enue associated with assigning a set of agents to a set of tasks.
Assignment problems with up to 30 agents and up to 30 tasks are
permitted.
Integer Linear Programming. This program will solve all integer and
mixed integer programming problems with up to 100 variables and 50
constraints. Both general integer and 0–1 variable problems can be
solved. Data input and editing features are the same as for the linear
programming module. Large numbers of integer variables significantly
increase solution time.
Shortest Route. This program will find the shortest route between any
two nodes of a network. Any node may be selected as the starting node
and any other node may be selected as the ending node. The program will
handle networks with a maximum of 20 nodes and a maximum of 40 arcs.
Minimal Spanning Tree. This program will find the minimal spanning
tree for a network that has a maximum of 20 nodes and a maximum of 40
arcs.
PERT/CPM. This program will develop activity schedules for projects with
up to 25 activities. Both the cases of certain activity times (single activity
time estimates) and uncertain activity times (three activity time
estimates) are allowed. The activity schedule, the critical activities, and
the expected project completion time are computed.
Inventory. This program provides six inventory models that can be used
to determine
the optimal order quantity, the optimal reorder point, and related
operating policies for a
3
variety of inventory situations. Both the cases of deterministic demand
and probabilistic demand are permitted.
Waiting Lines. This program will provide a summary of the operating

characteristics of
a variety of waiting line systems involving single and multiple channels.
Operating characteristics such as average waiting time, average number
of units in the system, probability of waiting, and so on are provided. An
economic analysis of the waiting line may be conducted.
Decision Analysis. This program will provide the best decision

alternative recommendation for problems having up to 10 decision
alternatives, up to 10 states of nature and known or unknown state of
nature probabilities. Up to 10 indicators may be considered in order to
develop a decision strategy based on sample information. The expected
values of perfect information and sample information are provided.
Forecasting. This program will compute forecasts for a time series by

using either moving averages, exponential smoothing, trend projection, or
trend and seasonal components. The mean square error measure of
forecasting accuracy and forecasts for future time periods are provided.
Time series with up to 100 data values or observations are permitted.
Markov Processes. This program will analyze Markov processes with up

to 10 states. Both Markov processes involving steady state behavior and
absorbing states can be analyzed.
1.4 USING THE MANAGEMENT SCIENTIST

This section describes the features of The Management Scientist that are
found in all 12 modules. An overview of this section is recommended
before you attempt to solve a problem using any of the modules.
Selecting a Module
After starting The Management Scientist, you will encounter the Select a
Module menu shown in Figure 1.1. The choices provide access to the 12
modules. Click the desired module title and a brief description of the
module will appear on the screen. To use the selected module simply click
OK or, alter natively, double click the module title.
Figure 1.1 The Select a Module Menu
4
Creating a New Problem
Whenever you want to begin working on a new problem, choose the
appropriate module and select the File pull-down menu. Then, selecting
the option New will begin the problem description and data input process
and allow you to create the new problem.
Saving a Problem
You may save a problem by selecting the File pull-down menu, and then
selecting the Save option. This will enable you to name the problem and
save it for later use. The Inventory and Waiting Line modules have a Save
button directly on the data input dialog box. The Save button will enable
you to save the problem without accessing the File pull-down menu.
After you specify the name of the problem, The Management Scientist
automatically adds a three-letter extension. The extension is used to
identify the module used to create the problem. For example, suppose you
have a linear programming problem that you want to save using the name
LPExample. The Management Scientist will add the extension LPM so that
your problem will be saved as a file LPExample.LPM. Only files with the
LPM extension can be saved and retrieved with the Linear Programming
module. Other modules have their own three-letter extensions. While you
can see the extensions on your directory of saved problems, you do not
5
have to input the extensions because they are handled automatically by
The Management Scientist.
Many users who elect to save problems find it a good file
management practice to create a separate directory for saved problems.
If you wish to do this, you will need to create your own problem directory.
Then, whenever you select the Save option, you will be able to designate
your problems directory as the place where you want to save the new
problem. If you name the problems directory Problems and if it is a
subdirectory of the directory containing The Management Scientist, The
Management Scientist will automatically go to the Problems directory
whenever you request Save.
Opening a Saved Problem

Suppose that you have previously saved a problem and that you now
want to open the problem file so that the problem can be modified and/or
solved again. Select the File pull-down menu and then select the Open
option. You will obtain a list of saved problems for the active module.
Select the problem you want to open and it will appear on your screen
ready for modification and/or solution.
Editing a Problem
Once you have created a new problem or opened a saved problem, you
may make modifications, changes, or corrections right on the screen. If
you have solved a problem and the solution is on the screen, you may go
back and edit the problem by selecting the Edit pull-down menu. Select-
ing the Display/Edit the Problem option will display the problem’s data
input screen. You can use this screen to make changes to the problem.
This editing process can be repeated any number of times, allowing you to
learn how the solution changes as various aspects of the problem are
modified.
For most of the modules, the Edit pull-down menu only has the
Display/Edit the Problem option. However, some of the modules have
an additional editing option. For example, the Linear Programming and
Integer Linear Programming modules have options on the Edit pull-down
menu that allow you to change the number of variables and/or the
number of constraints for the problem. The Transportation and
Assignment modules also have Edit options that enable you to change
the size of the problem.
Solving a Problem
When you have created a new problem, opened a saved problem, or
edited a current problem, you will most likely want to solve it. Select the
Solution pull-down menu. Then selecting Solve will solve the problem
and display the solution on the screen. The Inventory and Waiting Line
6
modules have a Solve button on the data input dialog box. In these
modules, the Solve button will enable you to display the solution without
accessing the Solution pull-down menu.
Printing a Solution
Once the solution appears on the screen, you may want to send it to a
printer or to a text file. This is accomplished by selecting the Solution
pull-down menu and then selecting the Print Solution option. The To
Printer or To Text File options indicate where the solution will be
directed. A text file can be later accessed by a word processor so that the
solution output can be edited and displayed as part of a solution report.
Changing Modules
After working with one module, you may want to change to a different
module and solve another type of problem. Selecting the File pull-down
menu and then selecting the Change Modules option will return you to
the Select a Module screen shown in Figure 1.1. At this point, you may
choose another module.
Exiting The Management Scientist

When you have completed a session with The Management Scientist,
select the File pull-down menu and then select the Exit option.
1.5 OTHER CONSIDERATIONS

We have covered most of the general features of The Management
Scientist. The following chapters contain more detailed information about
particular modules. Each module chapter provides one or more example
problems with a demonstration and discussion of data input, editing,
special features, and output. The following additional comments and
suggestions may be helpful as you use The Management Scientist.
Moving Around the Screen

In using The Management Scientist you will need to move around the
screen in order to select menus and interact with data input dialog boxes.
The mouse, the Tab key, and the arrow keys are the primary ways to
navigate The Management Scientist package. Using the mouse provides
the greatest degree of flexibility. Moving the mouse and clicking lets you
move to any menu or cell on the screen. During sessions with data input
and interactions with dialog boxes, the Tab key is very helpful. The Tab
sequence is designed to move you through the most logical sequence of
problem description and data input. After each piece of information is
7
entered, the Tab key will signal the computer to accept the cell’s input and
move the cursor automatically to the next logical step of the input
process. If the Tab key results in a move that is not where the user wants
to go, the mouse can be used to reposition the cursor. The arrow keys
may be used to navigate data input worksheets.
The mouse can be used to open the File, Edit, and Solution pull-down
menus and to select options from the menus. In addition to the mouse,
key strokes can be used. The underlined letter in the menu or option title
indicates the keystroke to use. Specifically, Alt-F will open the File pull-
down menu, Alt-E will open the Edit pull-down menu, and Alt-S will open
the Solution pull-down menu. Once a pull-down menu is opened, the Alt
key is no longer required and the underlined letter for the desired option
can be used. For example, in the File pull-down menu, N will activate
New, O will activate Open, S will activate Save, M will activate Change
Modules, and X will activate Exit.
Some Advice for Data Input

1.Do not enter commas (,) with your input data. For example, to enter
the numerical value of 104,000, simply type the six digits 104000.
2.Do not enter the dollar sign ($) for profit or cost data. For example,
a cost of $20.00 should be entered as 20.
3.Do not enter the percent sign (%) if percentage input is requested.
For a percentage of 25%, simply enter 25. Do not enter 25% or .25.
4.Occasionally a quantitative model will be formulated with fractional

values such as 1/4, 2/3, 5/6, and so on. The data input for the com-
puter must be in decimal form. The fraction of 1/4 can be entered
as .25. However, the fractions 2/3 and 5/6 have repeating decimal
forms. In cases such as these we recommend the convention of
rounding to 5 places. Thus, the corresponding decimal values of
.66667 and .83333 should be entered.
5.Finally, we recommend that in general you attempt to scale

extremely large or extremely small input data so that numbers of
roughly the same magnitude may be input and operated on by the
computer. For example, a cost of $2,500,000 may be scaled to 2.5
with the understanding that the data used in the problem reflect
millions of dollars.
Special Error Messages

All modules check input data for validity before proceeding. If data input is
not acceptable to the module, the computer will either not accept it or will
send a “beep” and a brief message describing the difficulty. In most cases
8
the messages will be self-explanatory and simply give you the opportunity
to re-enter corrected data values.
9
CHAPTER 2
LINEAR PROGRAMMING
Linear programming is one of the most widely used quantitative

approaches to decision making. Studies have indicated that most large
corporations make at least some use of linear programming. Applications
are numerous. Linear programming has been used to assist in production
scheduling, choose a port folio that will maximize return subject to
constraints on risk, solve a variety of resource allocation problems, and so
on.
A linear programming problem is a mathematical model of a decision
-making problem that can be solved to assist a manager in making better
decisions. The mathematical model is called a linear program because it
consists of a linear objective function and a set of linear constraints.
Solving the linear program means to optimize (maximize or minimize) the
objective function subject to the limitations and requirements represented
by the constraints.
To use a linear programming approach, a mathematical model that
describes the objective function of the problem as well as the constraints
must be formulated. Once this linear program has been developed, the
Linear Programming module of The Management Scientist can be used to
find the optimal solution.
2.1 OVERVIEW
The Linear Programming module uses the revised simplex method with LU
decomposition to solve linear programs containing up to 100 variables
and 50 constraints.
Significant features available with this module include:
1. The use of variable names of up to eight characters in length.
2. An on-screen editing capability that simplifies modifying problem

coefficients, adding and deleting variables, and so on.
3. A standard output that includes the optimal solution as well as sensitivity

analysis information. Dual prices and ranges for the objective function
coefficients and right-hand sides are provided.
4. Easy-to-use menus that simplify solving linear programming problems for
10
beginning students but still provide extensive problem-solving capabilities
for more experienced users.
5. The ability to save and retrieve problems from any disk drive and
subdirectory; when the user asks to retrieve a problem, a list of all pre-
viously saved problems is displayed.
6. An extensive error checking capability that prevents the user from getting
“hung up” because of an error in input.
2.2 AN EXAMPLE PROBLEM

Par, Inc., is a small manufacturer of golf equipment and supplies
concerned with developing a quarterly production plan for its standard
and deluxe golf bags. A careful analysis of manufacturing requirements
resulted in Table 2.1, which shows the production time requirements for
the four manufacturing operations and the accounting department’s
estimate of the profit contribution per bag.
Production Time (Hours)

Cutting Inspection Profit
and and per
Product Dyeing Sewing Finishing Packaging Bag
Standard 7/10 1/2 1 1/10 $10
Deluxe 1 5/6 2/3 1/4 $9
Table 2.1 Production Requirements and Profit Contributions per Bag
The director of manufacturing estimates that 630 hours of cutting and

dyeing time, 600 hours of sewing time, 708 hours of finishing time, and
135 hours of inspection and packaging time will be available for the
production of golf bags during the next three months. Par’s problem is to
determine how many standard and how many deluxe bags should be
produced in order to maximize profit contribution.
Letting X1 = number of standard bags Par, Inc., produces and X2 =
number of deluxe bags Par, Inc., produces, the following mathematical
model can be used to find the combination of X1 and X2 that satisfies all
the constraints and, at the same time, yields a value for the objective
function that is greater than or equal to the value given by any other
feasible solution.
max 10X + 9X
1 2
s.t.
7/10X + 1X
<_630 Cutting and Dyeing
1 2
1/2X +5/6X <_600 Sewing
1 2
11
1X +2/3X <_708 Finishing
1 2
1/10X +1/4X <_135 Inspection/Packaging
1 2
X , X >_ 0
1 2
2.3 CREATING AND SOLVING A PROBLEM

To solve the Par, Inc., problem, we begin by selecting the Linear
Programming module and choosing New from the File menu; the Problem
Features dialog box will then appear. Figure 2.1 shows the Problem
Features dialog box after entering 2 for the Number of Decision Variables,
4 for the Number of Constraints, and choosing Maximize for the
Optimization Type. After selecting OK, we obtain the data input screen
shown in Figure 2.2. Note that the two decision variables have initially
been given the names X1 and X2, and that the four constraints are
labeled Constraint 1, Constraint 2, Constraint 3, and Constraint 4.
Figure 2.1 Problem Features Dialog Box
Figure 2.2 Initial Data Input Screen
12
After entering the objective function coefficients, the constraint coeffi-
cients, the constraint relationships, and the right-hand side values, the
data input screen appears as shown in Figure 2.3. All fractional
coefficients must be converted to their decimal equivalent. For example,
the coefficient of X1 in the cutting and dyeing constraint must be written
as .7 instead of 7/10, the coefficient of X1 in the sewing constraint must
be written as .5 instead of 1/2, and so on. When rounding is necessary, we
recommend using the closest five-place decimal value; thus, the 5/6
coefficient of X2 in the sewing constraint is writ ten as .83333. Note also
that each of the four less-than-or-equal-to constraints is entered using <
instead of <_, and the nonnegativity constraints, X1, X2>_0 do not have
to be entered. After choosing Solve from the Solution menu, the optimal
solution will be presented as shown in Figure 2.4.
Figure 2.3 Data Input Screen for the Par, Inc., Problem
13
Objective Function Value = 7667.99417
Variable Value Reduced Costs
-------------- -------------- ---------------
X1 539.99842 0.00000
X2 252.00110 0.00000
Constraint Slack/Surplus Dual Prices

-------------- -------------- ---------------
1 0.00000 4.37496
2 120.00071 0.00000
3 0.00000 6.93753
4 17.99988 0.00000
OBJECTIVE COEFFICIENT RANGES

Variable Lower Limit Current Value Upper Limit
------------ -------------- ---------------- --------------
X1 6.30000 10.00000 13.49993
X2 6.66670 9.00000 14.28571
RIGHT HAND SIDE RANGES

Constraint Lower Limit Current Value Upper Limit
------------ -------------- ---------------- ---------------
14
1 495.60000 630.00000 682.36316
2 479.99929 600.00000 No Upper Limit
3 580.00140 708.00000 900.00000
4 117.00012 135.00000 No Upper Limit
Figure 2.4 Optimal Solution to Par, Inc., Problem
2.4 INTERPRETATION OF SOLUTION

First, note the number 7667.99417, which appears to the right of the
heading “Objective Function Value =.” Rounding this value, we can
conclude that the optimal solution to the Par, Inc., problem will provide a
profit of $7668. Directly below the objective function value we find the
values of the decision variables in the optimal solution. After rounding, we
have X1 = 540 standard bags and X2 = 252 deluxe bags as the optimal
production quantities.
The information in the column labeled “Reduced Costs” indicates how
much the objective function coefficient of each decision variable would
have to improve1 before it would be possible for that variable to assume a
positive value in the optimal solution. Thus, if a decision variable is
already positive— as is the case for both X1 and X2 in the Par, Inc.,
problem—its reduced cost is zero.
Immediately following the optimal X1 and X2 values and the reduced
cost information, the computer output provides information about the
status of the constraints. Since the Par, Inc., constraints are all of the less-
than-or-equal-to type, the information shown in the column labeled
“Slack/Surplus” provides the value of the slack variable for each of the
four constraints. This information is rounded and summarized below:
Constraint Slack
1 0
2 120
3 0
4 18
From the above information we see that the binding constraints (cutting
and dyeing and finishing constraints) have zero slack, or unused capacity,
the sewing department has 120 hours of unused capacity, and the
inspection and packaging department has 18 hours of slack, or unused
capacity.
The column in Figure 2.4 labeled “Dual Prices” contains information
about the marginal value of each of the four resources (time available in
each department) at the optimal solution. For The Management Scientist
the dual price is defined as follows:
1
For a maximization problem, improve means get bigger; for a minimization problem, improve
means get smaller.
15
The dual price associated with a constraint is the improvement in the
optimal value of the objective function per unit increase in the right-
hand side of the constraint.
Thus, the nonzero dual price of 4.37 for constraint 1 (cutting and
dyeing) and 6.94 for constraint 3 (finishing) tell us that an additional hour
of cutting and dyeing time will improve (increase) the value of the
objective function by $4.37 and an additional hour of finishing time will
improve (increase) the value of the objective function by $6.94. The range
over which this interpretation is applicable will be discussed shortly. Thus,
if the cutting and dyeing time were increased from 630 to 631 hours, with
all other coefficients in the problem remaining the same, Par’s profit would
be increased by $4.37 from $7668 to $7668 + $4.37 = $7672.37. Since
the sewing and the inspection and packaging constraints both have slack
or unused capacity available, the dual prices of zero show that additional
hours of these two resources will not improve the value of the objective
function.
From the information provided under the heading labeled OBJECTIVE
COEFFICIENT RANGES, we see that as long as the profit contribution
associated with the standard bag (currently a value of $10) is between
$6.30 and $13.50, the production of X1 = 540 standard bags and X2 = 252
deluxe bags will remain optimal. Similarly, as long as the profit
contribution associated with the deluxe bags is between $6.67 and
$14.29, the production of 540 standard bags and 252 deluxe bags will
remain optimal.
The final section of the computer output (RIGHT HAND SIDE RANGES)
contains ranging information for the constraint right-hand sides. As long
as the constraint right-hand-side value stays within this range, the
associated dual price gives the improvement in value of the objective
function per unit increase in the right-hand side. For example, since the
dual price for constraint 1 (cutting and dyeing) is $4.37, we can conclude
that additional hours will increase the objective function by $4.37 per hour
until the number of hours available reaches 682.36. It is also true that a
reduction in the hours available will reduce the value of the objective
function by $4.37 per hour. From the ranging information given we see
that the dual price of $4.37 is valid as long as the right-hand-side value is
between 495.60 and 682.36. For constraint 3 (finishing) we see that the
dual price of $6.94 is applicable as long as the right- hand-side value is
between 580 and 900.
At this point it is important to note that the sensitivity analysis (dual
prices and ranging information) presented in the computer output is
based on the assumption that coefficients are changed one at a time, with
all other coefficients of the problem remaining as stated in the original
problem. The sensitivity analysis information provided in the computer
output does not apply to two or more simultaneous changes in problem
coefficients.
16
As you solve more linear programming problems you will find that in many
cases only a few of the variables appear in each constraint (i.e., they have
coefficients of 0 in many constraints). The Management Scientist does not
require you to enter any terms that have 0 coefficients. This can
substantially reduce the time needed to enter a large problem.
It is often desirable to use more descriptive names for the decision
variables. For example, in the Par, Inc., problem, if you wanted to use the
name STANDARD instead of X1 and the name DELUXE instead of X2, you
enter the name STANDARD in the X1 variable names cell and DELUXE in
the X2 variable names cell. Variable names are restricted to any
combination of up to eight letters or numbers; the first character,
however, must be a letter.
When entering coefficients that are 1000 or greater, do not enter any
commas. For example, a constraint that required the variable named
FUNDSINV to be less than or equal to $100,000 would be written as
FUNDSINV < 100000
Note also that the $ symbol must be deleted when writing the right-hand-
side value.
When you enter a constraint make sure that no variables appear on
the right-hand side of the constraint. For example, if you have the
constraint
VARONE = VARTWO + 10
you must enter this constraint as
VARONE – VARTWO = 10
Only a nonnegative numerical value may be used on the right-hand side

of a constraint.
17
CHAPTER 3
TRANSPORTATION
PROBLEM
The transportation problem arises frequently in planning for the

distribution of goods and services from several supply locations to several
demand locations. Usually the amount of goods available at each supply
location (origin) is given and there is a specified amount needed at each
demand location (destination). With a variety of shipping routes and
differing costs for the routes, the objective is to determine how many
units should be shipped from each origin to each destination so that all
destination demands are satisfied and total transportation costs are
minimized. Other applications of the transportation problem involve
multiperiod production and inventory scheduling, cash management, and
a variety of network flow problems.
3.1 OVERVIEW
The Transportation module of The Management Scientist will find the
optimal solution to transportation problems with up to 30 origins and 30
destinations. The following input data must be entered by the user:
1.The number of origins.
2.The number of destinations.
3.The amount of supply at each origin.
4.The amount of demand at each destination.
5.The per unit cost or the per unit revenue of shipping from each
origin to each destination.

Foster Generators, Inc., produces a particular type of generator at plants
located in Cleveland, Ohio; Bedford, Indiana; and York, Pennsylvania.
Production capacities for these plants over the next three-month planning
18
period are as follows:
Three-Month
Production Capacity
Origin Plant (Units)
1 Cleveland 5000
2 Bedford 6000
3 York 2500
The firm distributes its generators through four regional distribution

centers located in Boston, Chicago, St. Louis, and Lexington; the three-
month forecast of demand at each distribution center is shown below:
Three-Month
Distribution
Demand Forecast
Destination Center (Units)
1 Boston 6000
2 Chicago 4000
3 St. Louis 2000
4 Lexington 1500
Management would like to determine the number of units to ship from

each plant to each distribution center. Since the production costs are the
same at each plant, the only variable costs involved are transportation
costs. Thus the problem becomes one of determining the distribution
routes to be used and the quantity to be shipped via each route so that all
distribution center demands can be met with a minimum total
transportation cost. The cost for each unit shipped on each route is given
in Table 3.1.
Destination
Origin Boston Chicago St. Louis Lexington
Cleveland 3 2 7 6
Bedford 7 5 2 3
York 2 5 4 5
Table 3.1 Transportation Cost per Unit for the Foster Generators Problem
19
Figure 3.1 Origins and Destinations Dialog Box

To solve the Foster Generators’ transportation problem, we begin by
selecting the Transportation module and choosing New from the File
menu; the Origins and Destinations dialog box will then appear. Figure 3.1
shows this dialog box after entering 3 for the number of origins and 4 for
the number of destinations. After selecting OK, we obtain the
Transportation Tableau data input screen shown in Figure 3.2.
Transportation cost per unit data from Table 3.1 are entered into the
corresponding cells of the Transportation Tableau. For example, Figure 3.2
shows a cost of $3 per unit for the origin 1 (Cleveland) to destination 1
(Boston) cell, $2 per unit for origin 1 (Cleveland) to destination 2 (Chi-
cago), and so on. After the cost per unit data are entered, the origin sup
plies are entered in the right-hand column of the tableau and destination
demands are entered in the bottom row of the tableau. When the data
input process is complete, choosing Solve from the Solution menu
provides a Select Optimization Criteria dialog box where the user may
specify whether a maximization objective or minimization objective is
desired. Selecting minimization for the Foster Generators and selecting OK
provides the optimal transportation solution as shown in Figure 3.3.
Recalling the following origin and destination numbers,
Origins Destinations
1. Cleveland 1. Boston
2. Bedford 2. Chicago
3. York 3. St. Louis
20
4. Lexington
the minimum cost solution calls for shipping 3500 units from Cleveland to
Boston, 1500 units from Cleveland to Chicago, 2500 units from Bedford to
Chicago, 2000 units from Bedford to St. Louis, 1500 units from Bedford to
Lexington, and 2500 units from York to Boston. The total transportation
cost of this solution is shown to be $39,500.
Figure 3.2 Transportation Tableau Data Input Screen
21
Figure 3.3 Solution Screen for the Foster Generators’ Problem

When supply is not equal to demand, the Transportation module
recognizes the condition automatically. If supply is greater than demand,
the solution will indicate which origins have excess supply. If demand
exceeds supply, the pro gram will find the best solution for the existing
supply; the destinations having unsatisfied demand are then displayed.
If there are origin–destination combinations that are unacceptable, you
must still enter a cost or revenue for each unacceptable combination. To
ensure that these unacceptable transportation routes are not included in
the optimal solution, enter a very large cost (e.g., 999999) or a very small
revenue (e.g., –999999) where appropriate.
The Transportation module provides the capability to add or delete ori-
gins and to add or delete destinations to an existing transportation
problem. Use the Edit menu to select the desired Add/Delete Origins or
Destinations. The Transportation Tableau will be displayed so that the
cost per unit and the supply or demand data may be entered.
Finally, if the user does not enter data in one or more cells of a Trans-
portation Tableau, The Management Scientist will assign 0 values to these
22
cells. When Solve is selected, The Management Scientist will provide a
warning message that some of the data cells have 0 entries. The user
may return to the Transportation Tableau and replace the 0 entries with
appropriate data or the user may solve the transportation problem with 0
entries assumed.
CHAPTER 4
ASSIGNMENT PROBLEM
The assignment problem arises in a variety of decision-making situations.

Typical assignment problems involve assigning jobs to machines,
assigning workers to tasks or projects, assigning sales personnel to sales
territories, assigning contracts to bidders, and so on. In general terms, the
assignment problem can be thought of as assigning agents to tasks. A
distinguishing feature of the assignment problem is that one agent is
assigned to one and only one task. The goal is to choose the set of
assignments that will optimize a stated objective, such as minimize cost,
minimize time, or maximize profits.
4.1 OVERVIEW
The Assignment module of The Management Scientist will find the optimal
solution to assignment problems with up to 30 agents and/or 30 tasks.
The following input data must be entered by the user:
1.The number of agents.
2.The number of tasks.
3.The cost or revenue associated with assigning each agent to each

task.

Fowle Marketing Research, Inc., is faced with the problem of assigning
three avail able project leaders (agents) to each of three new research
studies (tasks). Fowle’s management realizes, however, that the time
required to complete each study will depend on the experience and ability
of the project leader assigned to the study. Since the three studies have
been judged to have approximately the same priority, the company would
23
like to determine the assignments that will minimize the total number of
days required to complete all three. If a project leader is to be assigned to
one and only one client, what assignments should be made?
With three project leaders and three clients, there is a total of nine
possible assignment alternatives. The alternatives and the estimated
project completion times in days are summarized in Table 4.l.
Client
Project Leader 1 2 3
1 Terry 10 15 9
2 Carle 9 18 5
3 McClymonds 6 14 3
Table 4.1 Estimated Project Completion Times (Days) for the Fowle Assignment
Problem

To solve the Fowle Marketing Research assignment problem, we begin by
selecting the Assignment module and choosing New from the File menu;
the Agents and Tasks dialog box will then appear. Figure 4.1 shows this
dialog box after entering 3 for the number of agents and 3 for the number
of tasks. After selecting OK, we obtain the Assignment Tableau data
screen shown in Figure 4.2. Assignment time data from Table 4.1 are
entered into the corresponding cells of the Assignment Tableau. For
example, Figure 4.2 shows 10 days for the assignment of agent 1 (Terry)
to task 1 (Client 1), 15 days for the assignment of agent 1 (Terry) to task 2
(Client 2), and so on. When the data input process is completed, choosing
Solve from the Solution menu provides a Select Optimization Criteria
dialog box where the user may specify whether a maximization objective
or minimization objective is desired. Selecting minimization for the Fowle
Marketing Research problem and selecting OK provides the optimal
assignment solution as shown in Figure 4.3. Thus we see Agent 1 (Terry) is
assigned to Client 2, Agent 2 (Carle) is assigned to Client 3 and Agent 3
(McClymonds) is assigned to Client 1. The minimum total time required to
complete all three tasks is 26 days.
Figure 4.1 Agents and Tasks Dialog Box
24
25
Figure 4.2 Assignment Tableau Input Screen
26
Figure 4.3 Solution Screen for the Fowle Marketing Research Program

When the number of agents is not equal to the number of tasks, the
Assignment module recognizes the condition automatically. If the number
of agents is greater than the number of tasks, the solution will indicate
which agents are unassigned. If the number of tasks exceeds the number
of agents, the Assignment module finds the best solution for the existing
agents; the tasks that are not assigned an agent are then displayed.
If there are agent–task combinations that are unacceptable, you must
still enter a cost or revenue for each unacceptable agent–task
combination. To ensure that these unacceptable assignments are not
included in the optimal solution, enter a very large cost (e.g., 999999) or a
very small revenue (e.g., –999999) where appropriate.
The Assignment module provides the capability to add or delete
agents and to add or delete tasks to an existing assignment problem. Use
the Edit menu to select the desired Add/Delete Agents or Tasks. The
Assignment Tableau will be displayed so that the additional data may be
entered.
Finally, if the user does not enter data in one or more cells of an
Assignment Tableau, The Management Scientist will assign 0 values to
these cells. When Solve is selected, The Management Scientist will
provide a warning message that some of the data cells have 0 entries.
The user may return to the Assignment Tableau and replace the 0 entries
27
with appropriate data or the user may solve the assignment problem with
the 0 entries assumed.
28
CHAPTER 5
INTEGER LINEAR
PROGRAMMING
Problems that can be modeled as linear programs, except for the

additional requirement that some or all of the variables must be integers,
are called integer linear programming problems. The use of integer
variables provides additional modeling flexibility. As a result, the number
of practical applications that can be addressed with integer linear
programming methodology is quite large.
If all of the variables are required to be integers, we say we have an
all-integer linear program. For example, the following is an all-integer
linear programming model:
max 2X1 + 3X2

s.t.
3X1 + 3X2 <_ 12
2/3X1 + 1X2 <_ 4
1X1 + 2X2 <_ 6
X1, X2 >_ 0 and integer
Note that if the phrase “and integer” is dropped from the above model we
are left with the familiar two-variable linear program.
If some, but not all, of the variables in a problem are required to be
integers, we say we have a mixed-integer linear program. For example, in
the above model if the phrase “and integer” is replaced by “and X2
integer” we would have a mixed-integer linear program. In this case
although X2 would be restricted to only integer values, X1 would be said to
be continuous; it could assume any value greater-than-or-equal-to zero.
In most practical applications, the integer variables are only permitted
to assume the values 0 or 1. In such cases, we say that we have a binary,
or a 0–1, integer linear program. Zero–one problems may be either the
all-integer or mixed-integer type.
5.1 OVERVIEW
29
The Integer Linear Programming module of The Management Scientist
employs a branch and bound solution procedure. It can solve all-integer,
mixed-integer, and 0–1 integer linear programs with up to 100 variables
and 50 constraints. However, problems with large numbers of integer
variables may solve slowly. Just as with the Linear Programming module,
the number of decision variables that can be used must allow for slack,
surplus,
and artificial variables that are added automatically by the Integer Linear
Programming module.
Since the Linear Programming module and the Integer Linear
Programming module are the same with respect to creating, retrieving,
saving, and editing a problem, we will focus our attention on the
differences associated with solving integer linear programs. Readers
unfamiliar with the Linear Programming Module of The Management
Scientist should review Chapter 2 before continuing with this chapter.
5.2 AN EXAMPLE PROBLEM:

THE ALL-INTEGER CASE
Security Realty Investors currently has $1,365,000 available to purchase
town houses or apartment buildings. The townhouses can be purchased in
blocks of three for the price of $195,000 per block, but there are only four
blocks of town houses available for purchase at this time. Each building in
the apartment complex sells for $237,000, and the developer has agreed
to build as many units as Security would like to purchase.
Security’s property manager is free to devote 140 hours per month to
these investments. Each block of townhouses will require 4 hours of the
property manager’s time each month, while each apartment building will
require 40 hours per month. The yearly cash flow (after deducting
mortgage payments and operating expenses) is estimated at $2000 per
block of townhouses and $3000 per apartment building. Security would
like to allocate its investment funds to townhouses and apartment
buildings in order to maximize the yearly cash flow.
To develop an appropriate mathematical model, let us introduce the
following decision variables:
X1 = number of blocks of townhouses purchased

X2 = number of apartment buildings purchased
The objective function, measuring cash flow in thousands of dollars, can

be written as
max 2X1 + 3X2
There are three constraints that must be satisfied:
195X1 + 273X2 <_ 1365 Funds available ($1000s)
30
4X1 + 40X2 <_ 140 Manager’s time (hours)
X1 <_ 4 Townhouses available
Adding the nonnegativity requirements we obtain a linear programming
model involving the two variables and three constraints. This model could
be solved using the Linear Programming module described in Chapter 2;
the optimal linear programming solution is X1 = 2.442 and X2 = 3.256,
with an objective function value of 14.651.
However, since fractional values do not make any sense in the
context of the problem, the standard linear programming approach to this
problem is not appropriate. What we need to do is add the requirement
that both X1 and X2 are restricted to integer values. Doing this, we obtain
the following all-integer linear programming model:
max 2X1 + 3X2
s.t.
195X1 + 273X2 <_ 1365 Funds available ($1000s)
4X1 + 40X2 <_ 140 Manager’s time (hours)
X1 <_ 4 Townhouses available
X1, X2 > _ 0 and integer

To solve the Security Realty Investors problem, we begin by selecting the
Integer Linear Programming module and choosing New from the File
menu; the Problem Features dialog box will then appear. Figure 5.1 shows
the Problem Features Dialog box after entering 2 for the Number of
Decision Variables, 3 for the Number of Constraints, and choosing
Maximize for the Optimization Type. After selecting OK, we obtain the
data input screen shown in Figure 5.2. Note that the two decision
variables have initially been given the names X1 and X2, and that the
three constraints are labeled Constraint 1, Constraint 2, and Constraint 3.
Figure 5.1 Problem Features Dialog Box
31
32
After entering the objective function coefficients, the constraint coeffi-

cients, the constraint relationships, and the right-hand-side values, we
obtain the data input screen shown in Figure 5.3. Note that each of the
three less-than-or-equal-to constraints is entered using < instead of <_,
and the nonnegativity constraints, X1, X2 >_ 0, do not have to be entered.
After choosing Solve from the Solution menu, we obtain the Integer
Variable Identification dialog box shown in Figure 5.4. After selecting the
All Variables Are General Integer option and choosing Solve, the optimal
solution will be presented as shown in Figure 5.5.
33
Figure 5.3 Data Input Screen for the Security Realty Investors Problem
Figure 5.4 Integer Variable Identification Dialog Box
34
Figure 5.5 Optimal Solution to the Security Realty Problem
We see that the optimal solution for Security Realty is to purchase

four blocks of townhouses (X1 = 4) and two apartment buildings (X2 = 2);
this solution corresponds to a yearly cash flow of $14,000.
Note that the form of the optimal solution output is similar to that of
the Linear Programming module. However, the reduced costs and dual
prices are not given. This is because they do not have the same meaning
as for linear programming problems. Also, no ranging information is given
for objective function coefficients and right-hand sides. Such information
is not meaningful for integer linear programs.

THE 0–1 ALL-INTEGER CASE
The Ice-Cold Refrigerator Company can invest funds over the next four
years in a variety of company projects that have differing capital
requirements. Faced with limited funds, the company must select the
35
most profitable projects and budget for the necessary capital
expenditures. The estimated present values for the projects, the capital
requirements, and the available capital projections are as follows:
Estimated
Present Capital Requirements
Project Value($) Year l Year 2 Year 3 Year 4
Plant expansion 90,000 15,000 20,000 20,000 15,000
Warehouse expansion 40,000 10,000 15,000 20,000 5,000
New machinery 10,000 10,000 0 0 4,000
New product research 37,000 15,000 10,000 10,000 10,000
Available capital funds 40,000 50,000 40,000 35,000
The following decision variable definitions are used:
X1 = 1 if the plant expansion project is accepted;

0 if rejected
X2 = 1 if the warehouse expansion project is accepted;

0 if rejected
X3 = 1 if the new machinery project is accepted;

0 if rejected
X4 = 1 if the new product research project is accepted;

0 if rejected
Using these decision variable definitions, the objective function can be

written as shown below (monetary values are expressed in thousands of
dollars):
max 90X1 + 40X2 + 10X3 + 37X4
The mathematical model of this capital budgeting problem has a separate

constraint for each year’s available funds; thus we obtain the following
four constraints:
15X1 + 10X2 + 10X3 + 15X4 <_ 40

20X1 + 15X2 10X4 <_ 50
20X1 + 20X2 10X4 <_ 40
15X1 + 5X2 + 4X3 + 10X4 <_ 35
X1, X2, X3, X4 = 0,1
To solve the above integer linear programming problem using the
36
Integer Linear Programming module, we enter the problem in the usual
fashion. After verifying that the problem has been entered correctly, we
choose Solve from the Solution menu. Then, in the Integer Variable
Identification dialog box, we select the All Variables Are 0/1 (Binary) option
and choose Solve; the optimal solution, shown in Figure 5.6, is given by X1
= 1, X2 = 1, X3 = 1, and X4 = 0 with a total estimated present value of
$140,000.
Figure 5.6 Optimal Solution to the Ice-Cold Refrigerator Company

Problem

THE MIXED-INTEGER CASE
Suppose that we had the following mixed-integer linear programming
problem:
max 1X1 + 1X2

s.t.
7X1 + 9X2 <_ 63
9X1 + 5X2 <_ 45
3X1 + 1X2 <_ 12
X1, X2 >_ 0 and X2 integer
37
After entering this problem, we choose Solve from the Solution menu.
Then, in the Integer Variable Identification dialog box we select the
Multiple Types of Variables option. From the General Integer Variables list
we choose X2.
After selecting Solve, the optimal mixed integer solution is displayed
as shown in Figure 5.7.
Figure 5.7 Optimal Solution to the Mixed-Integer Example Problem

The Integer Linear Programming module of The Management Scientist is a
general integer programming system; that is, it has the capability of
solving every type of integer linear programming problem. A branch and
bound solution procedure, which requires the solution of a linear pro-
gramming problem at each node, is employed. The number of
computations involved in solving these problems can be substantial for
even moderately sized problems. Thus, solution times may take as long as
several minutes for more difficult problems and will depend on the speed
of your personal computer.
Sensitivity analysis is often more critical for integer linear
programming problems than for linear programming problems. A very
small change in one of the coefficients in the constraints can cause a
relatively large change in the value of the optimal solution. But, the dual
38
prices and ranging information made available with the solution of linear
programming problems are not helpful with integer linear programming
problems. Thus, practitioners usually recommend resolving an integer
linear program several times with slight variations in the coefficients to
explore sensitivity issues before attempting to choose an optimal solution
for implementation.
39
CHAPTER 6
SHORTEST ROUTE
In this chapter, we consider the network application of designing a

transportation system where the primary interest is in determining the
shortest route or shortest path through the network. These types of
problems arise in situations where it is desired to transport goods
between a variety of locations and the objective is to find the most
economical route between various pairs of locations.
The cost of transporting goods from one location to another is usually
measured in distance traveled, but travel time or dollar cost could be used
just as easily. Other situations in which the need to solve a shortest route
problem might arise, include the design of computer networks, vehicle
routing problems, and so on.
6.1 OVERVIEW
The Shortest Route module of The Management Scientist will find the
shortest distance between any 2 nodes in a network consisting of up to 20
nodes and 40 arcs. To create and solve a new problem, the user must first
develop a net work diagram of the problem to be solved and then provide
the following data input to the Shortest Route module:
1.The number of nodes
2.The number of arcs
3.For each arc:

a. The number of the start node
b. The number of the end node
c. The distance (or other measure used) between the nodes
4.The 2 nodes it is desired to find the shortest route between

The Gorman Construction Company operates several construction projects
40
located throughout a three-county area. With multiple daily trips carrying
personnel, equipment, and supplies to and from the construction sites, the
costs associated with transportation activities are substantial. The
network shown in Figure 6.1 describes the travel alternatives to and from
six of Gorman’s newest construction sites. The circles (nodes) of the
network correspond to the site locations. The lines between the nodes
(arcs) correspond to the roads and the streets between the locations. The
numbers next to each arc represent road distances in miles. If Gorman
wishes to minimize the travel distance from the office to site 7, what is
the route that should be followed? Note that the net work has 7 nodes and
10 arcs.

To solve the Gorman shortest route problem, we begin by selecting the
Shortest Route module and choosing New from the File menu; the Nodes
and Arcs dialog box will then appear. Figure 6.2 shows this dialog box after
entering 7 for the number of nodes and 10 for the number of arcs. After
selecting OK, we obtain the Network Description data input screen shown
in Figure 6.3. Net work description data are entered into the
corresponding cells. For example, row 1 corresponds to arc 1 of the
network, which starts at node 1, ends at node 2, and has a distance of 15
miles. Row 2 corresponds to arc 2 of the network, which starts at node 1,
ends at node 3, and has a distance of 10 miles. The numbering of the 10
arcs as shown in Figure 6.3 is arbitrary. A different user might number the
arcs differently and input the network description data in a different order.
The Management Scientist is flexible and can accept the network arc
descriptions in any order. When the data input process is completed,
choosing Solve from the Solution menu provides a Select the Starting and
Ending Nodes dialog box where the user may specify any starting node
and any ending node in the network. Selecting starting node 1 and ending
41
node 7 for the Gorman problem and selecting OK provides the optimal
shortest route as shown in Figure 6.4. Thus we see that the shortest route
between nodes 1 and 7 passes through nodes 3, 5, and 6. The total
distance is 22 miles.
42
Figure 6.2 Nodes and Arcs Dialog Box
Figure 6.3 Network Description Data Input Screen
43
Figure 6.4 Shortest Route Between Nodes 1 and 7 for the Gorman
Problem
At this point, the user may select Solve from the Solution menu again.
Altering the starting node and/or ending node and selecting OK will
provide the shortest route between the two specified nodes. This process
can be repeated in order to find the shortest route between any other pair
of nodes in the network.
44
CHAPTER 7
MINIMAL
SPANNING TREE
The minimal spanning tree problem involves using the arcs of the network
to connect all nodes in such a fashion that the total length of the arcs
used is minimized. As with the shortest route problem, it is customary to
use distance as a measure of arc length. However, in some applications it
is more appropriate to use arc cost, time to go from one node to another,
and so on.
A common situation in which the minimal spanning tree problem arises
involves the design of communication networks, where the objective is to
minimize the amount of the cable necessary to connect all nodes that
must communicate with one another. Other applications involve designing
road networks, developing the layout for conveyor systems, and so on.
7.1 OVERVIEW
The Minimal Spanning Tree module of The Management Scientist will find
the optimal solution to minimal spanning tree problems with up to 20
nodes and 40 arcs. The module will find the set of arcs that creates a path
connecting every node to every other node in such a manner that the
total length of all arcs used is minimized. To create and solve a new
problem, the user must first develop a network diagram of the problem to
be solved and then provide the following input data:
1.The number of nodes
2.The number of arcs
3.For each arc:

a. The number of the start node
b.The number of the end node
c. The distance (or other measure used) between the nodes
45
The Southwestern Regional Computer Center must have special computer
communication lines installed in order to connect five satellite users with
a new central computer. Because the lines are expensive, the computer
center’s management wants the total length of the new communication
lines to be as small as possible. While the central computer could be
connected directly to each user, it appears to be more economical to
install a direct line to some users and let other users tap into the system
by linking up with the users who are already connected to the system. The
network shown in Figure 7.1 shows us the links being considered between
the regional computer center and the satellite users. The distance, in
miles, between locations is shown above the associated arcs. Note that
the network has 6 nodes and 11 arcs.

To solve the Regional Computer System minimal spanning tree problem,
we begin by selecting the Minimal Spanning Tree module and choosing
New from the File menu; the Nodes and Arcs dialog box will then appear.
Figure 7.2 shows this dialog box after entering 6 for the number of nodes
and 11 for the number of arcs. After selecting OK, we obtain the Network
Description data input screen shown in Figure 7.3. Network description
data are entered into the corresponding cells. For example, row 1
corresponds to arc 1 of the net work, which starts at node 1, ends at node
2, and has a distance of 20 miles. Row 2 corresponds to arc 2 of the
network, which starts at node 1, ends at node 3, and has a distance of 40
miles. The numbering of the 11 arcs as shown in Figure 7.3 is arbitrary. A
different user might number the arcs differently and input the network
description data in a different order. The Management Scientist is flexible
and can accept the network arc descriptions in any order. When the data
46
input process is completed, choosing Solve from the Solution menu
provides the minimal spanning tree solution as shown in Figure 7.4. Thus
we see that communication lines must be installed between five pairs of
locations: 1 and 2, 1 and 4, 4 and 3, 4 and 6, and 3 and 5. The total length
of the communication lines is 110 miles.
47
Figure 7.2 Nodes and Arcs Dialog Box
Figure 7.3 Network Description Data Input Screen
48
Figure 7.4 Minimal Spanning Tree for the Regional Computer Center
Problem
49
CHAPTER 8
PERT/CPM
PERT (Program Evaluation and Review Technique) and CPM (Critical Path
Method) are approaches management scientists have developed to assist
managers with planning, scheduling, and controlling projects. PERT and
CPM were initially developed independently for slightly different purposes,
but in today’s usage, the distinction between PERT and CPM as two
separate techniques has largely disappeared.
PERT/CPM has been used to plan, schedule, and control a wide variety
of projects, such as:
1.Research and development of new products and processes
2.Construction of plants, buildings, and highways
3.Maintenance of large and complex equipment
4.Design and installation of new systems
In projects such as the above, project managers must schedule and

coordinate the various activities so that the entire project is completed on
time. A complicating factor in carrying out this task is the
interdependence of the activities; some activities depend on the
completion of other activities before they can be started. The PERT/CPM
technique can assist managers by helping them answer questions such as
the following:
1.What is the time required to complete the project?
2.What start and finish dates should be scheduled for each activity?
3.Which are the critical activities?
4.How long can noncritical activities be delayed before they become

critical?
8.1 OVERVIEW
50
The PERT/CPM module of The Management Scientist can be used to
analyze projects with up to 25 activities. The program will develop an
activity schedule, identify the critical activities, and compute the expected
project completion time. It can be used for projects with known activity
completion times and for projects with uncertain activity completion
times. For projects with uncertain activity completion times, the PERT/CPM
module will compute the expected time and variance for each activity as
well as the expected completion time and variance for the entire project.
Before using the PERT/CPM module, the user must develop a list of the
activities and, for each activity, the immediate predecessors. Each activity
must be identified by a letter of the alphabet and no activity may have as
an immediate predecessor an activity identified by a letter coming later in
the alphabet. The data input requirements differ depending on whether
activity times are known or uncertain. We illustrate the use of the module
in both cases.
8.2 AN EXAMPLE PROBLEM WITH

KNOWN ACTIVITY COMPLETION TIMES
The owner of the Western Hills Shopping Center is considering
modernizing and expanding his current 32-store shopping complex. The
activities necessary to complete the expansion project are listed in Table
8.1. Note that the list includes the immediate predecessor for each
activity as well as the number of weeks required to complete the activity.
Completion
Activity Immediate Time
Activity Description Predecessor (Weeks)
A Prepare architectural drawings — 5
B Identify potential tenants — 6
C Develop prospectus A 4
D Select contractor A 3
E Prepare building permits A 1
F Secure approval of building permits E
4
G Construction D, F 14
H Finalize tenant contracts B, C 12
I Tenants move in G, H 2
Table 8.1 Activity List for the Western Hills Shopping Center Expansion
Project
Information in Table 8.1 indicates that the total time required to

complete all activities in the shopping center expansion project is 51
weeks. However, from the immediate predecessor information, we see
that several of the activities can be completed simultaneously (A and B,
51
for example). Being able to work on two or more activities at the
same time will shorten the total project completion time to less than 51
weeks. We now show how the PERT/CPM module can be used to provide
information in managing this project.

To analyze the Western Hills Shopping Center expansion project, we begin
by selecting the PERT/CPM module and choosing New from the File menu;
the PERT/CPM Parameters dialog box will then appear. Figure 8.1 shows
the PERT/CPM Parameters dialog box after selecting the Known Activity
Times option and entering 9 for the Number of Activities. After selecting
OK, we obtain the data input screen shown in Figure 8.2.
Figure 8.1 PERT/CPM Parameters Dialog Box
52
For each of the nine activities in the Select Activity Name list we must
identify each activity’s immediate predecessor and the expected activity
time for each activity. For example, we begin by clicking the letter A in the
Select Activity Name list. Since A does not have any immediate
predecessor, we only need to enter the expected completion time of 5 in
the Expected Time box. The input for other activities follows in a similar
fashion, with the
exception that for activities with immediate predecessors, we must also
choose the immediate predecessors from the Select Predecessor list.
Figure 8.3 shows the data input screen after entering all the data for the
Western Hills Shopping Center expansion project. After choosing Solve
from the Solution menu, the activity schedule, critical path, and project
completion time will be displayed as shown in Figure 8.4.
Activities A–E–F–G–I are on the critical path and the project may be
completed in 26 weeks. The earliest start, latest start, earliest finish,
latest finish, and slack times are shown for each activity.
Figure 8.3 Data Input Screen for the Western Hills Shopping Center
Project
53
54
Figure 8.4 Activity Schedule for the Western Hills Shopping Center
Project
8.4 AN EXAMPLE PROBLEM WITH UNCERTAIN

ACTIVITY COMPLETION TIMES
The H.S. Daugherty Company is considering manufacturing a cordless
vacuum cleaner that can be powered by a rechargeable battery. The
vacuum cleaner, referred to as a Porta-Vac, is expected to contribute to
Daugherty’s expansion into the household market. Daugherty’s
management has initiated a project to study the feasibility of proceeding
with the Porta-Vac idea. The list of activities and immediate predecessors
are shown in Table 8.2 .
Activity Description Immediate Predecessors

A Develop product design —
B Plan market research —
C Routing (manufacturing engineering) A
D Build prototype model A
E Prepare marketing brochure A
F Cost estimates C
G Preliminary product testing D
H Market survey B, E
I Pricing and forecast report H
J Final report F, G, I
55
Table 8.2 Activity List for the Daugherty Porta-Vac Project
For the Daugherty Porta-Vac project, activity times are uncertain. In

such cases, the PERT procedure calls for three time estimates (optimistic,
most probable, and pessimistic) to be used for each activity. The activity
time estimates for the Porta-Vac project are given in Table 8.3.
Most
Optimistic Probable Pessimistic
Activity Time Time Time
A 4 5 12
B 1 1.5 5
C 2 3 4
D 3 4 11
E 2 3 4
F 1.5 2 2.5
G 1.5 3 4.5
H 2.5 3.5 7.5
I 1.5 2 2.5
J 1 2 3
Table 8.3 Optimistic, Most Probable, and Pessimistic Activity Time

Estimates (in weeks) for the Porta-Vac Project
To analyze the Daugherty Porta-Vac project using the PERT/CPM

module, we begin by choosing New from the File menu. In the Enter PERT
/CPM Parameters dialog box, we select the Uncertain Activity times option,
enter 10 for the Number of Activities, and select OK; the problem input
screen will then appear. Then, for each activity, in addition to the
information on immediate predecessors, we must enter the optimistic,
most probable, and pessimistic time estimates; Figure 8.5 shows the
resulting problem input screen.
After choosing Solve from the Solution menu, we obtain the output
shown in Figure 8.6. The PERT/CPM module prints a list of the expected
times and variances for activities, followed by the activity schedule, the
critical path, the expected project completion time, and the variance of
project completion time. For the Daugherty Porta-Vac project the critical
path is A–E–H–I–J, with an expected project completion time of 17 weeks.
The variance of project completion time is shown to be 2.72.
56
Figure 8.5 Problem Input Screen for the Daugherty Porta-Vac Project
57
Figure 8.6 Activity Schedule for the Daugherty Porta-Vac Project
58
CHAPTER 9
INVENTORY
Inventories can be defined as any idle products or materials that are

waiting to be used. For most companies, the expenses associated with
financing and maintaining inventories are a substantial part of the cost of
doing business.
Quantitative models of inventory systems have been developed to
help managers make decisions such as the following:
1.How much should be ordered when the inventory for a particular

item is to be replenished?
2.When should the inventory of the item be replenished?
In the best known and most fundamental inventory model, referred to as

the economic order quantity model, the objective is to satisfy demand
while minimizing inventory holding costs and ordering costs. In other
inventory models, backorder costs, quantity discount costs, and stockout
costs are also considered.
While there are many similarities in all inventory systems, each
system has unique characteristics that prevent the application of one or
two general inventory models to all situations. For example, in some
inventory systems it may be reasonable to assume that demand is
deterministic; that is, the demand for the item occurs at a known and
constant rate. In other inventory systems, however, the demand may be
probabilistic; that is, the demand fluctuates and can only be described in
probabilistic terms.
Based on assumptions about the type of demand encountered and
given relevant costs, inventory models can be used to provide information
such as the following:
1.Minimum cost order quantity
2.Annual inventory holding cost
3.Annual order cost
4.Total annual cost
59
5.Maximum inventory level
6.Average inventory level
7.Reorder point
8.Number of orders per year
9.Cycle time (time between orders)
Using the information provided by an inventory model, managers are

better equipped to establish a low-cost inventory policy.
9.1 OVERVIEW
The Inventory module of The Management Scientist provides the following
six inventory models:
1.Economic Order Quantity—Given a constant demand rate, this

model determines the order quantity that will minimize the total
annual inventory holding and ordering costs.
2.Economic Production Lot Size—Given a constant demand rate and a

constant production rate, this model determines the production lot
size that will minimize the total annual inventory holding and
production setup costs.
3.Economic Order Quantity with Planned Shortages—This model

allows planned shortages or backorders. Given a constant demand
rate, this model determines the order quantity that will minimize
the total annual inventory holding, ordering, and backordering
costs.
4.Economic Order Quantity with Quantity Discounts—This model

considers cases where the purchase cost per unit changes
depending on the order quantity. Given a constant demand rate,
this model deter mines the order quantity that will minimize the
total annual inventory holding, ordering, and purchasing costs.
5.Order Quantity–Reorder Point with Probabilistic Demand—This

model follows the approach of the economic order quantity model
with the addition that demand during the lead time period is
probabilistic and described by a normal probability distribution. The
model determines the order quantity and reorder point that will
minimize annual inventory holding and ordering costs and provides
an acceptable value for the probability of a stockout.
60
6.Single-Period Inventory with Probabilistic Demand—With demand
described by a uniform or normal probability distribution, this model
determines the optimal order quantity for single-period inventory
situations in which items cannot be carried in inventory from one
period to the next.

R&B Beverage Company is a distributor of beer, wine, and soft drink
products. For each of its products, R&B would like to determine how much
to order and when to order so that inventory holding and ordering costs
are minimized. To illustrate how the Inventory module can help in
determining answers to these questions, we will apply the economic order
quantity model to R&B’s best selling product, Bub Beer.
The annual demand for Bub Beer is assumed to be approximately con-
stant at 104,000 cases. It costs $32.00 to place an order with the supplier,
and each case of Bub costs R&B $8.00. The annual inventory holding cost
per unit is 25% of the cost of the item, and the R&B Beverage Company
operates 250 days per year. In addition, when a new order is placed with
the supplier, it takes two days for the order to reach the R&B warehouse.
Given these data regarding Bub Beer, management would like to answer
the following questions:
1.What is the minimum cost order quantity of Bub Beer?
2.What is the reorder point or inventory level that signals R&B should
order a new shipment of Bub Beer?
3.How many orders will be placed per year and what is the length of
time between orders?
4.What is the total annual inventory holding and ordering cost for this
product?

To solve the Bub Beer inventory problem, we begin by selecting the
Inventory module and choosing New from the File menu; the Inventory
Models Avail able dialog box shown in Figure 9.1 will then appear. This
dialog box can be used to select any one of the six inventory models
listed. Since the Bub Beer problem is an economic order quantity
application, we select the economic order quantity model. The dialog box
shown in Figure 9.2 appears on the screen. The data for the Bub Beer
problem are entered as shown. Note that the Holding Cost Method shows
Percent with the Percent of Unit Cost = 25 (for 25%) and Unit Cost = 8.
The Holding Cost Method can be changed to Amount, which requires the
61
user to input the holding cost amount per unit. The Percent and Amount
options enable inventory holding cost to be expressed as either a
percentage of the unit cost or as a specific dollar amount. The Number of
Working Days per Year will show 250 days, which is a common
assumption. If 250 days per year is not appropriate, this entry can be
changed to reflect the actual number of working days per year for the
problem. Finally, Compute Reorder Point appears as a question since the
economic order quantity model can be implemented with or without
reorder point information. Since the Bub Beer problem requested reorder
point information, the Compute Reorder Point box has been clicked on and
the lead time of 2 days has been entered.
62
Figure 9.1 Inventory Models Available Dialog Box
Figure 9.2 Economic Order Quantity Dialog Box
After the data input process is completed, Solve or Save can be imple-
mented by selecting the appropriate button in the data input dialog box.
Selecting Solve provides the optimal solution shown in Figure 9.3. Thus,
we see that the economic order quantity for Bub Beer is 1824.28 units,
and the reorder point is 832 units. This optimal inventory policy will result
in approximately 57 orders per year with the time between orders 4.39
days. Note also that the total annual cost is $3,648.56.
63
Figure 9.3 Output for the R&B Beverage Economic Order Quantity
Problem
It is well known that the optimal order quantity provided by the eco-
nomic order quantity model is not sensitive to rounding error. Thus, in
practice, rounding the optimal order quantity to a more convenient
quantity (such as 1800, 1900, or 2000 units) can be expected to have
little effect on the total annual cost.
9.4 OTHER INVENTORY MODELS

There are six different inventory models that may be solved using this mo-
dule (see Figure 9.1). The process of creating and solving a problem is
similar for all the models. Here, we briefly overview the differences in data
input and the output provided.
Economic Production Lot Size

This model is appropriate for problems involving a production run. It is
similar to the economic order quantity model; the production lot size run
64
corresponds to the economic order quantity and the production setup cost
corresponds to an order cost. Attention is usually focused on the following
questions:
1.What is the minimum cost production lot size?
2.What is the reorder point or inventory level that signals a new

production run should be scheduled?
3.How many production setups will be scheduled per year and what is
the length of time between production runs?
4.What is the total annual inventory holding and setup cost for this
product?
Data input for the economic production lot size model is very similar
to the data input for the economic order quantity model shown in Figure
9.2. One additional input item—the annual production rate—is required for
the production lot size model.
Economic Order Quantity with Planned Shortages

This inventory model uses the assumption of the economic order quantity
model with the exception that it allows planned shortages called
backorders. The model assumes that if the item is out of inventory, the
customer requesting the item will wait until the next incoming order is
received, and such back orders will be filled as soon as the new order is
received.
Data input includes the economic order quantity data input shown in
Figure 9.2 plus one additional piece of information: the cost per unit per
year for a backorder. Output
information includes the standard economic order quantity information
shown in Figure 9.3. In addition, the planned stockouts model indicates
the maximum number of back orders that will occur, the annual cost of
the backorders, and the total cost including inventory holding, ordering,
and backordering costs.
Economic Order Quantity with Quantity Discounts

This model is based on the economic order quantity model with the
variation that purchase cost per unit is a relevant cost and depends on the
order quantity decision. Thus, in this model it may be desirable to order
additional units in order to obtain a quantity discount, which will result in
a lower unit cost.
The economic order quantity with quantity discounts model requires
the same input data as the economic order quantity model (see Figure
9.2). In addition, this model requires you to enter quantity discount
65
information showing the unit cost and the minimum quantity to order to
obtain this cost for each discount category.
The output information is the same as for the economic order quantity
model as shown in Figure 9.3, with the addition that purchase cost is
included in the total cost. The recommended optimal order quantity will
be in the discount category providing the lowest total cost.
Order Quantity–Reorder Point with Probabilistic

Demand
This inventory model is another variation of the basic economic order
quantity model. The difference between the two models is that the order
quantity– reorder point with probabilistic demand model assumes that
demand during the lead time is subject to uncertainty. A normal
probability distribution is used to describe the lead time demand.
Data input includes the basic economic order quantity data input
shown in Figure 9.2, with the exception that the reorder point is based on
the input of the mean and standard deviation of the lead time demand
rather than on the input of lead time in days. The computation of the
reorder point is made based on a desired service level. The service level
can be expressed in terms of either
1.An allowable number of stockouts per year, or
2.The probability of a stockout during an inventory cycle.
The choice of the expression of service level is up to the user. Under

option 1, the allowable number of stockouts per year is specified. Under
option 2, the acceptable probability of a stockout during an inventory
cycle is specified. Generally, most inventory systems will seek to have
inventory avail able for customers and thus the allowable number of
stockouts per year and the probability of a stockout per cycle will be low.
The output information displayed includes the economic order
quantity model information shown in Figure 9.3. In addition the output
displays the safety stock level, the safety stock cost, the expected
number of stockouts per year, and the probability of a stockout per cycle.
Single-Period Inventory with Probabilistic Demand

This inventory model considers probabilistic demand for the situation
where inventory cannot be carried from one period to the next. In a single-
period inventory situation, one order is placed for the item; at the end of
the period the item has either sold out or there is a surplus of unsold
items, which will be sold for a salvage value. The single-period model
applies to situations involving seasonal or perishable items that cannot be
carried in inventory and sold in future periods.
The following three input values are required to use this model:
66
1.The cost of overestimating demand and ordering too much
2.The cost of underestimating demand and not ordering enough
3.The probability distribution for demand, which must be either a uni-

form or a normal probability distribution
The output information shows the optimal order quantity, the

probability of a surplus, and the probability of a stockout.
67
CHAPTER 10
WAITING LINES
Waiting lines occur in many everyday situations. People wait in

supermarket check-out lines, at bank teller windows, in doctor offices, at
gasoline service stations, at airline ticket counters, and so on. In general,
units (customers) arrive at a service facility seeking the service provided
by the facility. If the service facility is busy, the units form a waiting line
and wait until the service facility is available.
Quantitative models have been developed that enable decision
makers to identify the operating characteristics of waiting line systems.
Typical information provided by waiting line models includes the following:
1.The probability that the service facility is idle and there are no units
in the system
2.The average number of units in the waiting line
3.The average number of units in the system (units waiting plus units
being serviced)
4.The average time a unit spends in the waiting line
5.The average time a unit spends in the system (waiting time plus
service time)
6.The probability that an arriving unit has to wait for service
7.The total system cost (the cost of idle time for the units in the
system plus the cost of operating the service facility)
Given the above information, decision makers are better able to make
waiting line design and policy decisions that balance the desired service
levels with the cost of providing the service.
10.1 OVERVIEW
The Waiting Lines module of The Management Scientist provides the
following five waiting line models.
68
1.Single- or Multiple-Channel with Poisson Arrivals Exponential
Service Times—This is the fundamental waiting line model with
Poisson arrivals and exponential service times. Operating
characteristics are provided for systems having from 1 to 25
channels.
2.Single-Channel with Poisson Arrivals and Arbitrary Service Times—

This model is restricted to a single-channel system. However,
service times may follow any probability distribution. The only
information needed about the service times is the mean service
rate and the standard deviation of the service times.
3.Single-Channel with Poisson Arrivals and Deterministic Service

Times—
This model considers deterministic (constant) service times for a
single-channel system. All units require the same amount of time;
no variation in service time is permitted.
4.Single- or Multiple-Channels with Poisson Arrivals, Arbitrary Service

Times, and No Waiting Line—This model considers situations where
waiting is not permitted. An arriving unit either receives service
immediately or is blocked from the system. Blocked units are not
permitted to wait, are cleared from the system, and may try to re-
enter at a later point in time. Operating characteristics are provided
for systems having from 1 to 25 channels.
5.Single- or Multiple-Channels with Poisson Arrivals, Exponential

Service Times, and a Finite Calling Population—In this waiting line
model, a finite population of units is permitted to arrive for service.
Populations with a maximum of 30 units are allowed. Operating
characteristics are pro vided for systems having from 1 to 25
channels.

The Burger Dome fast-food restaurant operates a single-channel service
facility for its customers. A single Burger Dome employee takes the
customer’s order, determines the total cost of the order, takes the money
from the customer, and then fills the order. Once a customer’s order is
filled, the employee takes the order from the next customer waiting for
service. The Poisson probability distribution describes the customer
arrivals. The mean arrival rate is 0.75 customers per minute. The
exponential probability distribution describes the service times. The mean
service rate is one customer per minute. For economic considerations,
Burger Dome evaluates customer waiting time at $10 per hour while the
service channel can be operated for $7 per hour.
69
Management is interested in developing a better understanding of the
operating characteristics of the food-service waiting line. Specifically,
management would like to know the probability that no customers are in
the system, the average number of customers waiting for service, the
average time a customer waits for service, the probability that an arriving
customer has to wait, and the hourly cost of the service operation.

To solve the Burger Dome waiting line problem, we begin by selecting the
Waiting Lines module and choosing New from the File menu; the Waiting
Lines dialog box shown in Figure 10.1 will then appear. This dialog box can
be used to select any one of the five waiting line models listed. Since the
Burger Dome problem has a waiting line with Poisson arrivals and
exponential service times, we select the Poisson Arrivals/Exponential
Service model. The Poisson Arrivals/Exponential Service dialog box shown
in Figure 10.2 appears on the screen. The data for the Burger Dome
problem are entered with Number of Channels = 1, Mean Arrival Rate =
.75, and Mean Service Rate = 1. Economic Analysis appears as a question
since the waiting line model can be implemented with or without an
economic analysis. Since the Burger Dome problem requested an
economic analysis, the Economic Analysis box has been clicked on and
the customer waiting time of $10 per hour and service channel cost of $7
per hour have been entered.
Figure 10.1 Waiting Lines Dialog Box
Figure 10.2 Poisson Arrivals/Exponential Service Dialog Box
70
After the data input process is completed, Solve or Save can be imple-
mented by selecting the appropriate button in the data input dialog box.
Selecting Solve provides the output shown in Figure 10.3. While no direct
recommendation is made, the output provides valuable information about
the operating characteristics of the waiting line. An average of 2.25
customers are in the waiting line, the customers wait an average of three
minutes for service, and .75 or 75% of the arriving customers have to wait
for service. The total cost is $37.00 per hour. In addition, the probabilities
of the number of customers in the system show that some long waiting
lines are possible. The information in Figure 10.3 indicates that something
should be done to improve the operation of the Burger Dome service
facility.
Figure 10.3 Output for the Burger Dome Waiting Line Problem
71
Using the Edit menu, we can easily display the data input dialog box
again and change the number of channels from 1 to 2. Selecting Solve
shows the improved operating characteristics for the expanded Burger
Dome service operation. The total cost of the two-channel system is
$22.73 per hour. Repeating the edit process shows that the addition of a
third channel with the total cost increasing to $28.65 per hour is not
necessary.

Each of the five models in the Waiting Lines module requires the input
data to be based on the same period of time. The user may select the
most convenient period of time such as a minute, hour, day, and so on.
However, once selected, all input data must be stated in terms of the
selected period of time. The output information will appear in terms of the
selected time period. For example, if the user selects an hourly period of
time, a mean arrival rate of 45 customers per hour and a mean service
rate of 60 customers per hour will result in hourly output waiting times.
On the other hand, if the user wanted to use a minute as the time period,
the mean arrival rate would be converted to 45/60 = 0.75 customers per
minute and the mean service rate would be converted to 60/60 = 1
customer per minute; in this case the output waiting times will be in
minutes.
Model 2 allows the user to specify any arbitrary service time and
requires care to ensure that proper units are maintained during the data
input process. The data input requires the mean arrival rate, the mean
service rate, and the standard deviation of service times to be stated in
terms of the same period of time. For example, assume that data for a
particular application have a mean arrival rate of two units per hour,
service times with a mean of 20 minutes, and a standard deviation of 12
minutes. Assume that an hour is the time period selected by the user. The
mean arrival rate is the stated two units per hour. However, on an hourly
basis the mean service rate is 60/20 = 3 units per hour. The standard
deviation, which is input directly as time, must be converted to the same
hourly basis. Thus the standard deviation of service times would be 12/60
= 0.20 hour. If all the input is not stated
in terms of the same unit of time, the module will provide incorrect and
misleading output information.
72
CHAPTER 11
DECISION ANALYSIS
Decision analysis can be used to help a decision maker select the best
decision alternative for problems involving several decision alternatives
and an uncertain or risk-filled pattern of future events. A complicating
factor in selecting the best alternative is that the future events, called
states of nature, are uncontrollable; thus, the decision maker must select
a decision alternative before knowing which state of nature will occur.
The following three steps can be used to structure a decision analysis
problem:
1.Identify the decision alternatives
2.Identify the states of nature
3.Identify the payoff associated with each decision alternative and

each state of nature combination
The above information can be summarized in a payoff table; the decision

alternatives correspond to the rows of the table, the states of nature
correspond to the columns of the table, and the payoffs are the values in
each row and column.
The decision analysis approach to selecting the best decision
alternative depends on whether or not probabilities are specified for the
states of nature. If probabilities for the states of nature are not specified,
the decision maker first selects one of several decision criteria; then, a
decision alternative is recommended based on the criteria preference of
the decision maker.
If probabilities for the states of nature are specified, decision analysis
can be used to provide the following information:
1.The expected value or long-run average payoff for each decision

alternative
2.A recommended decision alternative based on the best expected

value
3.Expected value of perfect information—the expected value if the
73
decision maker has the perfect information (i.e., the decision maker
knows in advance which state of nature will occur)
In addition, decision analysis with probabilities provides the option of

using sample information to revise the probabilities of the states of
nature. In many cases a decision strategy based on the sample
information can provide a better expected value.
11.1 OVERVIEW
The Decision Analysis module of The Management Scientist is designed to
provide a recommended decision alternative for problems having the
following characteristics:
1.Up to 10 decision alternatives
2.Up to 10 states of nature
3.Known or unknown state of nature probabilities
A variety of measures such as profit, cost, time, utility, and so on may be

used for the payoff.
For decision making without probabilities, the decision maker has the
choice of three decision criteria: optimistic, conservative, and minimax
regret. The module will provide a recommended decision based on the
criterion selected.
For decision making with probabilities, the module will provide a
recommended decision based on the expected value criterion. Information
is also provided for the expected value of perfect information.
When probabilities are used, an option is also provided for developing
a decision strategy based on sample information. In this situation, sample
information will guide the decision maker in selecting the best decision
alternative. The module will identify the optimal decision strategy, its
expected value, the expected value of the sample information, and the
efficiency of the sample information.

Political Systems, Inc., (PSI) is a computer service firm specializing in
information services such as surveys and data analysis for individuals
running for political office. PSI is in the final stages of selecting a
computer system for its Midwest office. The three decision alternatives PSI
is considering are as follows:
d1 = a large computer system
74
d2 = a medium-sized computer system
d3 = a small computer system
The uncertainty of the customer acceptance of the PSI service makes

it difficult to know for sure which computer system the company should
select. The following two states of nature for acceptance have been
identified:
s1 = high customer acceptance of the PSI services
s2 = low customer acceptance of the PSI services
For now, assume that probability information is not available for the states
of nature.
The payoff table showing the projected profit for each decision
alternative and each state of nature combination is shown in Table 11.1.
Given this information, which computer system should PSI select?
States of Nature
Decision Alternatives High Acceptance Low Acceptance
Large Computer System 200,000 –20,000
Medium-Sized Computer System 150,000 20,000
Small Computer System 100,000 60,000
Table 11.1 Payoff Table for the PSI Problem

To solve the PSI decision analysis problem, we begin by selecting the
Decision Analysis module and choosing New from the File menu; the
Problem Features dialog box will then appear. Figure 11.1 shows this
dialog box after entering 3 for the Number of Decision Alternatives and 2
for the Number of States of Nature. For now, we will leave the State of
Nature Probabilities turned off and analyze the PSI problem without using
probability information. In Section 11.4 we will discuss the PSI problem
when probability information is used.
Figure 11.1 Problems Features Dialog Box
75
After selecting OK in the Problem Features dialog box of Figure 11.1,
we obtain the Payoff Table data input screen shown in Figure 11.2. Payoff
data from Table 11.1 are entered into the corresponding cells of the Payoff
Table as shown in Figure 11.2. When the payoff data input process is
completed, choosing Solve from the Solution menu provides the Select
Optimization/Decision Criteria dialog box shown in Figure 11.3. For the PSI
problem, we selected maximize the payoff as the optimization procedure
and optimistic as the decision criterion. Selecting OK provides the output
shown in Figure 11.4. The optimistic criterion recommends decision
alternative 1, the large computer system.
Figure 11.2 Data Input Screen for the PSI Problem
76
Figure 11.3 Select Optimization/Decision Criteria Dialog Box
Figure 11.4 Output Information for the PSI Problem Using the Optimistic
Criterion
77
11.4 USING PROBABILITY INFORMATION
Let us assume that state of nature probabilities are available for the PSI
problem with the probability of high customer acceptance (state of nature
1) 0.3 and the probability of low customer acceptance (state of nature 2)
0.7. Decision analysis uses the probability information and the expected
value criterion to make a decision recommendation.
With probability information available, The Management Scientist will
consider this a new problem. Use the File menu and select New to obtain
the Problem Features dialog box as shown in Figure 11.1. This time click
the State of Nature Probabilities box. Selecting OK provides the Payoff
Table data input screen shown in Figure 11.2 along with two additional
cells used to enter the two state of nature probabilities. When the data
input process is completed, choosing Solve from the Solution menu
provides the Select Optimization Criteria dialog box, which enables the
selection of a maximize or minimize criterion. Selecting maximize and
then selecting OK provides the output shown in Figure 11.5. In this case,
decision alternative 3, the small computer system, is recommended. The
expected value is $72,000. It is also shown that the expected value of
perfect information is 30,000.
78
Figure 11.5 Output Information for the PSI Problem Using the Expected
Value Criterion
Sample Information and a Decision Strategy

If you are using state of nature probabilities, you will be given the option
of selecting whether or not you would like to use sample information to
develop a decision strategy. Sample information usually takes the form of
research or experimentation, which helps to better access the
probabilities of the states of nature. Thus prior to selecting a decision
alternative, sample information will be collected. This sample information
will enable the revision of the state of nature probabilities. Ultimately, an
optimal decision strategy will be computed; the strategy will show how
the recommended decision alternative should be changed depending on
the sample information.
Let us return to the PSI problem and assume that a marketing
research study will be undertaken in order to obtain sample information
about possible customer acceptance of the PSI service. The study is
referred to as an indicator, and the results of the study are referred to as
indicator outcomes. The two indicator outcomes for the PSI problem are as
follows:
I1 = favorable market research report (individuals contacted generally

express considerable interest in PSI’s services)
79
I2 = unfavorable market research report (individuals contacted
generally express little interest in PSI’s services)
In general, the use of sample information requires the identification of

two or more indicator outcomes. In addition to specifying the indicator
outcomes, we must also have information about the conditional
probability of each indicator outcome for each state of nature. The
conditional probability information for the PSI problem is shown in Table
11.2. For example, given state of nature 1 were to occur, there is a 0.8
probability the indicator outcome will be favorable and a 0.2 probability
the indicator outcome will be
unfavorable.
Indicator Outcome
States of Nature Favorable (I1) Unfavorable (I2)
High acceptance (s1) 0.8 0.2

Low acceptance (s2) 0.1 0.9
Table 11.2 Conditional Probabilities of Indicator Outcomes for Each State

of Nature
Let us now consider how the decision analysis module can be used to
develop an optimal decision strategy based on the results of the market
research indicator outcomes. In addition, we will determine the expected
value of the sample information provided by the market research study.
When a decision analysis problem requires the computation of a
decision strategy, the Problem Features dialog box appears as shown in
Figure 11.6. Note that both State of Nature Probabilities and Compute
Decision Strategy have been clicked on. The values shown for the number
of decision alternatives, number of states of nature, and number of
indicator outcomes are for the PSI decision strategy problem.
Figure 11.6 Problem Features Dialog Box for the PSI Decision Strategy
Problem
80
Selecting OK provides the Payoff Table data input screen shown in
Figure 11.7. The top part of the screen provides the payoff table input as
before. The sections labeled Enter State of Nature Probabilities and
Indicator Probabilities are used to input the state of nature probabilities
and the indicator probabilities from Table 11.2. When the data input
process is completed, choosing Solve from the Solution menu provides the
Select Optimization Criteria dialog box, which enables the selection of a
maximize or minimize criterion. Selecting maximize and then selecting OK
provides the output shown in Figure 11.8. The optimal decision strategy
shows decision alternative d1 (large computer system) if the indicator
outcome is I1 (favorable market research report), and decision alternative
d3 (small computer system) if the indicator outcome is I2 (unfavorable
market research report).
Figure 11.7 Data Input Screen for the PSI Decision Strategy Problem
81
82
Figure 11.8 Optimal Decision Strategy for PSI
Additional output information shows that if indicator outcome I1 occurs

the expected profit will be $150,322.58. The probability of this indicator
outcome is 0.31. Also note that if indicator outcome I2 occurs, the
expected profit will be $63,478.26. The probability of this indicator
outcome is 0.69. The expected value of the decision strategy using the
indicator outcome information is $90,400. Thus the expected value of the
market research sample information is $18,400. The value of sample
information is 61.3% of the value that could be obtained with perfect
information.
83
CHAPTER 12
FORECASTING
A critical aspect of managing any organization is planning for the future.

Indeed, the long-run success of an organization is closely related to how
well management is able to foresee the future and develop appropriate
strategies. Forecasting techniques are available to help managers predict
many future aspects of a business operation.
Historical data on quantities such as sales, demand, stock prices,
university enrollments, and so on are collected over time and form what
are called time series. Specifically, a time series is a set of observations
measured at successive points in time or over successive periods of time.
The manager who wishes to forecast a particular time series will find that
the historical values of the time series provide clues about future values.
Quantitative forecasting methods are based on an analysis of
historical data. When the historical data used is restricted to past values
in the time series being forecasted, the forecasting procedure is called a
time series method.
12.1 OVERVIEW
The Forecasting module of The Management Scientist can be used to
analyze time series data consisting of up to 100 time periods or
observations. Future values of the time series may be forecast by any of
the following methods:
1.Moving Averages—This method uses the average of the most recent

data values to forecast the next value for the time series. The
number of data values in the moving average forecast must be
specified by the user.
2.Exponential Smoothing—This method uses a combination of the

forecast for the most recent period and the actual time series value
for the most recent period to forecast the next value for the time
series. The weight used for the most recent time series value is
referred to as the smoothing constant and must be specified by the
user.
3.Trend Projection—This method uses the criterion of least squares to
84
develop a linear trend equation, which expresses the time series
values as a linear function of the time period. Forecasts are made
by evaluating the linear trend equation for future time periods.
4.Trend and Seasonal Components—This method applies to a time

series that has both trend and seasonality. First the seasonal
influence is removed from the time series by a process referred to
as deseasonalizing the data. Then the criterion of least squares is
used to develop a linear trend equation, which expresses the
deseasonalized time series as a linear function of the time period.
The linear trend equation is used to forecast future values for the
deseasonalized time series. Finally, an adjustment for the
seasonality of the time period provides the forecast for the time
series.
A given time series may be analyzed by one or more of the above

fore casting methods. A comparison of forecasting accuracy for the
methods will indicate which method appears to have the best potential for
forecasting future values of the time series.

FORECASTING USING MOVING AVERAGES
Consider the 12 weeks of data presented in Table 12.1. These data
provide a time series of the number of gallons of gasoline sold by a
gasoline distributor in Bennington, Vermont, over the past 12 weeks. Use
the method of moving averages to forecast the number of gallons of
gasoline that will be sold in week 13.
Sales
Week (1000s of Gallons)
1 17
2 21
3 19
4 23
5 18
6 16
7 20
8 18
9 22
10 20
11 15
12 22
Table 12.1 A Time Series of Gasoline Sales
85
To forecast the number of gallons of gasoline that will be sold in week 13
for the time series shown in Table 12.1, we begin by selecting the
Forecasting module and choosing New from the File menu. When the Time
Periods dialog box appears, enter 12 for the Number of Time Periods and
choose OK; these steps result in the data input screen. Figure 12.1 shows
the data input screen after entering the gasoline sales time series data.
After selecting Solve from the Solution menu, we obtain the Forecasting
Methods dialog box shown in Figure 12.2. Any one of the four methods
listed may be used to forecast values for the time series.
Figure 12.1 Data Input Screen for the Gasoline Sales Time Series
Figure 12.2 Forecasting Methods Dialog Box
86
If Moving Averages is selected, you must enter the Number of Periods
to be included in the moving average. While any number of periods from 1
to the number of periods in the time series (12 in this example) may be
used in the moving average, values of 3 to 5 tend to be most common. In
order to continue this example, we used 3 as the number of periods to be
included in the moving average.
After you specify the number of periods for the moving average and click
Solve, the Forecasting module provides a summary of the three-week
moving average calculations for the gasoline time series; Figure 12.3
shows the output provided. The time series values, the moving averages
forecasts, and the forecast errors are shown for the entire time series.
Since an important consideration in using any fore casting method is the
accuracy of the forecasts, the mean square error (MSE), an often used
measure of the accuracy of a forecasting method, is also provided. The
forecasting method providing the lowest mean square error has provided
the best forecasts for the historical time series values. Also shown is the
fore cast for the next period in the time series. Thus, we see that the
three-period moving averages forecast for week 13 is 19,000 gallons.
Figure 12.3 The Forecasting Error and Forecast for the 3-Week Moving
Averages
87
12.4 OTHER FORECASTING METHODS
The Forecasting module contains four methods that may be used to
develop forecasts for a time series. These methods are made available by
the Forecasting Methods dialog box shown in Figure 12.2. An example of
the moving aver ages method was presented in the preceding section. In
this section we provide an overview of each of the other three methods.
We will describe the method and then discuss the data input and output
information features that differ from the moving averages example in
Section 12.3.
Exponential Smoothing
Exponential smoothing appears as choice 2 in the Forecasting Methods
dialog box in Figure 12.2. This method uses a combination of the forecast
for the most recent period and the actual time series value for the most
recent period to forecast the next value for the time series. The weight
used for the most recent time series value is referred to as the smoothing
constant. If you select the exponential smoothing method, you will be
asked to enter the value of the smoothing constant. This value must be
between 0 and 1.
Output information includes the time series values, the exponential
smoothing forecast, and the forecast errors for the historical time series.
In addition, the mean square error and the forecast for the next time
period are also provided.
88
Trend Projection
Trend projection appears as choice 3 in the Forecasting Methods dialog
box shown in Figure 12.2. This method uses the criterion of least squares
to develop a linear trend equation, which expresses the time series values
as a linear function of the time period. The trend projection analysis
begins as soon as this option is selected. No additional input information
is required.
The output information will provide a linear trend equation in the form
T = b0 + b1t
where
t = the time period
T = the trend value of the time series in period t
b0 = the intercept of the trend line
b1 = the slope of the trend line
The numerical values of b0 and b1 will be provided as part of the output.

Additional output information is similar to the output provided by the
moving averages and exponential smoothing routines.
Trend and Seasonal Components

Forecasting with trend and seasonal components appears as choice 4 in
the Forecasting Methods dialog box shown in Figure 12.2. This method
should be used for time series that are believed to have both linear trend
and seasonality. If you select this forecasting method you will be asked to
enter the number of seasonal indexes to be computed. In order to do this
you must have an idea of how many seasons are present in the time
series. The forecasting method will develop a seasonal index for each
season present. If you have a time series with monthly data for several
years, you will probably request 12 seasonal indexes to consider the
possibility that each month has a different seasonal effect. If you have a
time series with quarterly data for several years, you will probably request
4 seasonal indexes to consider the possibility that each quarter has a
different seasonal effect.
After you enter the number of seasonal indexes, the output will
appear on the screen in three separate sections. The first section will
provide the seasonal indexes. Figure 12.4 shows an example of this output
for a time series with quarterly data. The fourth quarter with an index of
1.141 shows that the time series is 114.1% higher in this quarter when
the seasonal influence is considered. The second quarter with an index of
0.836 shows that the time series will tend to be lower than normal in this
quarter. The other output sections follow the format of Figure 12.3; that is,
the past time series values, the fore casts, the forecast errors, the mean
89
square error, and the forecasts for each of the next three time periods are
provided, as well as an option of forecasting additional periods in the
future. Simply request this option, enter the time period you wish to
forecast, and the Forecasting module will provide the forecast.
FORECASTING WITH TREND AND SEASONAL COMPONENTS
SEASON SEASONAL INDEX
1 0.931
2 0.836
3 1.092
4 1.141
Figure 12.4 An Example with Four Seasonal Indexes
90
CHAPTER 13
MARKOV PROCESSES
The analysis of brand loyalty and brand switching behavior is one of the
major applications of Markov processes. In this application, the purchase
decision that is made each time period by a customer involves selecting
one of several competing brands. The brands are referred to as the states
of the process. Given that the customer has purchased a specific brand
during the current time period, transition probabilities describe the
probabilities that the customer will purchase the same brand as well as
each of the competing brands during the following time period.
The transition probability for a given state to itself is a measure of
brand loyalty in that it indicates the probability that the customer will
purchase the same brand in two successive periods. The transition
probabilities for a given state to each of the other states measure the
brand switching behaviors of the customer. Based on known or assumed
transition probabilities, a Markov process model can be used to answer
questions such as the following:
1.In the long run, what market share will each brand have?
2.If a new and/or improved brand enters the market, what impact can
be expected on the market share of each brand?
3.For a given group of customers, how many will purchase each brand
each time period?
Markov process models have also been developed for machine break
down behavior, estimating allowances for bad debt, and university
enrollment projections. In each instance it is assumed that (1) there are a
finite number of states, (2) the transition probabilities remain constant
over time, and (3) the probability of being in a particular state at any one
time period depends only on the state of the process during the preceding
time period.
13.1 OVERVIEW
The Markov processes module of The Management Scientist will analyze
problems with up to 10 states. Input to the program is the matrix of
91
transition probabilities for the states. The solution provides the steady-
state probabilities for the states.
In some applications it may not be possible to make a transition out of
one or more states once the state has been reached. Such states are
referred to as absorbing states. The Markov processes module will solve
problems for which the total number of absorbing and nonabsorbing
states is 10 or less. In applications with absorbing states, the solution
provides the probability that units currently in each of the nonabsorbing
states will eventually end up in each of the absorbing states.

Two grocery stores in a small town compete for customers. Each customer
makes one shopping trip per week to one of the two stores. A survey of
store loyalty among the customers shows that for customers who shop at
Murphy’s Foodliner one week, 90% will shop at Murphy’s the next week
and 10% will switch to Ashley’s Supermarket. For customers who shop at
Ashley’s Super market one week, 20% will switch to Murphy’s the next
week and 80% will shop again at Ashley’s. These transition probabilities
are summarized as follows:
Current Weekly Next Weekly Shopping Period

Shopping Period Murphy’s Foodliner Ashley’s Supermarket
Murphy’s Foodliner 0.9 0.1
Ashley’s Supermarket 0.2 0.8
What are the steady-state probabilities for the two grocery stores? If there
are 1,000 customers that make weekly shopping trips to one of the two
stores, how many customers can be expected to shop at each of the
stores?

To determine the steady-state probabilities for the two grocery stores in
our example problem, we begin by selecting the Markov processes
module and choosing New from the File menu. When the Markov
Processes dialog box appears, enter 2 for the Number of States and
choose OK; the Transition Matrix data input screen will then appear. Figure
13.1 shows the Transition Matrix after entering the transition probabilities
for the grocery store example problem. After selecting Solve from the
Solution menu, we obtain the output shown in Figure 13.2.
As you can see, Murphy’s Foodliner (state 1) has the higher steady-
state probability. Thus, we conclude that in the long run, Murphy’s
92
Foodliner will have a 66.7% share of the market and Ashley’s Supermarket
will have the remaining 33.3% share of the market. With 1,000 weekly
customers, 667 should shop at Murphy’s and 333 should shop at Ashley’s.
Figure 13.1 Transition Matrix Data Input Screen
Figure 13.2 Output for the Grocery Store Market Share Problem
93
WITH ABSORBING STATES
Heidman’s Department Stores has two aging categories for its accounts
receivable: (1) accounts that are classified as 0 to 30 days old and (2)
accounts that are classified as 31 to 90 days old. If any portion of an
account balance exceeds 90 days, that portion is written off as a bad
debt. The total account balance for each customer is placed in the age
category corresponding to the oldest unpaid amount; hence, this method
of aging accounts receivable is called the total balance method.
Let us assume that Heidman’s shows a total of $3,000 in its accounts
receivable and the firm’s management would like an estimate of how
much of the $3,000 will eventually be collected and how much will result
in bad debts.
To see how we can view the accounts receivable operation as a
Markov process, consider what happens to one dollar currently in accounts
receivable. As the firm continues to operate into the future, we can
consider each week as a trial of a Markov process with a dollar existing in
one of the following states of the system:
State Description
1 Paid category
94
2 Bad debt category
3 0 to 30 days age category
4 31 to 90 days age category
Based on historical transitions of accounts receivable dollars, the following

transition probabilities were developed for Heidman’s Department Stores:
To State
From State 1 2 3 4
1 1.0 0.0 0.0 0.0
2 0.0 1.0 0.0 0.0
3 0.4 0.0 0.3 0.3
4 0.4 0.2 0.3 0.1
The fact that the transition probabilities from state 1 to state 1 and from
state 2 to state 2 are both 1 shows that these two states are absorbing
states. That is, once a dollar is paid (state 1) it is always paid. Similarly,
once a dollar is declared a bad debt (state 2), it remains a bad debt. This
leads us to conclude that all accounts receivable dollars will eventually be
absorbed into either the paid or the bad debt state, and hence the name
absorbing state.
Management of Heidman’s Department Stores would like to answer
the following questions:
1.What is the probability that dollars currently in the 0 to 30 day age

category will eventually be paid?
2.What is the probability that dollars currently in the 31 to 90 day age

category will eventually be paid?
3.If $1,000 is currently in the 0 to 30 day category and $2,000 is

currently in the 31 to 90 day category, how much will be paid and
how much will end up as bad debt?
The problem creation step for this problem is the same as that
described previously. No special input procedures are required to handle
the absorbing states. The output is shown in Figure 13.3. There is a .889
probability that the 0 to 30 day age category dollars will be paid and a
.741 probability that the 31 to 90 day age category dollars will be paid.
Thus, of the current $3,000 of account receivables, the amount paid is
projected to be .889($1,000) + .741($2,000) = $2,371. An allowance for
bad debt should be set up for $3,000 – $2,371 = $639.
Figure 13.3 Output for the Accounts Receivable Problem
95
96

Management Scientist User's Guide

Uploaded by

Document Informationclick to expand document informationThis user's manual accompanies version 5. Of The Management Scientist. It was developed by David R. Anderson, Dennis J. Sweeney, and Thomas A. Williams. The new version 5.0 is available FOR WINDOWS 95 AND WINDOWS 98 environments.

Document Informationclick to expand document information

Copyright:

Available Formats

Management Scientist User's Guide

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Management Scientist User's Guide

Uploaded by

Copyright:

Available Formats

THE MANAGEMENT SCIENTIST

COPYRIGHT ©2003 by South-Western, a division of Thomson Learning.

Chapter 2 Linear Programming

Chapter 3 Transportation Problem

Chapter 4 Assignment Problem

Chapter 5 Integer Linear Programming

Chapter 6 Shortest Route

Chapter 7 Minimal Spanning Tree

Chapter 10 Waiting Lines

Chapter 11 Decision Analysis

Chapter 13 Markov Processes

Dr. David R. Anderson

Welcome to version 5.0 of The Management Scientist. You are ready to

You do not have to be an expert in quantitative analysis to use this

1.1 USING THIS MANUAL

2.Continue to read, or at least skim, the rest of Chapter 1 to obtain an

3.Select any one of the 12 modules of The Management Scientist to

1.Place the disk in the CD-ROM drive.

2.In Windows from the Start taskbar, choose Run.

4.Click the OK button or press the Enter key.

5.Follow the prompts of the installation.

Running The Management Scientist

1.3 OVERVIEW OF THE SOFTWARE PACKAGE

Modules Available in The Management Scientist

Transportation. This program will find the minimum cost or maximum

Waiting Lines. This program will provide a summary of the operating

Decision Analysis. This program will provide the best decision

Forecasting. This program will compute forecasts for a time series by

Markov Processes. This program will analyze Markov processes with up

1.4 USING THE MANAGEMENT SCIENTIST

Figure 1.1 The Select a Module Menu

Opening a Saved Problem

Exiting The Management Scientist

1.5 OTHER CONSIDERATIONS

Moving Around the Screen

Some Advice for Data Input

4.Occasionally a quantitative model will be formulated with fractional

5.Finally, we recommend that in general you attempt to scale

Special Error Messages

Linear programming is one of the most widely used quantitative

1. The use of variable names of up to eight characters in length.

2. An on-screen editing capability that simplifies modifying problem

3. A standard output that includes the optimal solution as well as sensitivity

4. Easy-to-use menus that simplify solving linear programming problems for

2.2 AN EXAMPLE PROBLEM

Production Time (Hours)

Table 2.1 Production Requirements and Profit Contributions per Bag

The director of manufacturing estimates that 630 hours of cutting and

2.3 CREATING AND SOLVING A PROBLEM

Figure 2.1 Problem Features Dialog Box

Figure 2.2 Initial Data Input Screen

Constraint Slack/Surplus Dual Prices

OBJECTIVE COEFFICIENT RANGES

RIGHT HAND SIDE RANGES

Figure 2.4 Optimal Solution to Par, Inc., Problem

2.4 INTERPRETATION OF SOLUTION

FUNDSINV < 100000

you must enter this constraint as