Académique Documents
Professionnel Documents
Culture Documents
Quantitative Analysis
for Management
Barry Render
Ralph M. Stair, Jr.
MODULES
M1
M2
M3
M4
M5
M6
E d i t i o n
M O D U L E
Statistical
Quality
Control
LEARNING OBJECTIVES
After completing this module, students will be able to:
1. Define the quality of a product or service.
2. Develop four types of control charts: x , R, p, and c.
3. Understand the basic theoretical underpinnings of statistical quality control,
including the central limit theorem.
MODULE OUTLINE
M1.1
M1.2
M1.3
M1.4
M1.5
Introduction
Defining Quality and TQM
Statistical Process Control
Control Charts for Variables
Control Charts for Attributes
Summary Glossary Key Equations Solved Problems Self-Test
Discussion Questions and Problems Case Study: Bayfield Mud Company
Case Study: Morristown Daily Tribune Bibliography
M1-1
M1-2
Module 1
M1.1 INTRODUCTION
For almost every product or service, there is more than one organization trying to make a
sale. Price may be a major issue in whether a sale is made or lost, but another factor is
quality. In fact, quality is often the major issue; and poor quality can be very expensive for
both the producing firm and the customer.
Consequently, firms employ quality management tactics. Quality management, or as
it is more commonly called, quality control (QC), is critical throughout the organization.
One of the managers major roles is to ensure that his or her firm can deliver a quality
product at the right place, at the right time, and at the right price. Quality is not just of
concern for manufactured products either; it is also important in services, from banking to
hospital care to education.
We begin this module with an attempt to define just what quality really is. Then we
deal with the most important statistical methodology for quality management: statistical
process control (SPC). SPC is the application of the statistical tools we discussed in Chapter 2 to the control of processes that result in products or services.
To some people, a high-quality product is one that is stronger, will last longer, is built
heavier, and is, in general, more durable than other products. In some cases this is a good
definition of a quality product, but not always. A good circuit breaker, for example, is not
one that lasts longer during periods of high current or voltage. So the quality of a product
or service is the degree to which the product or service meets specifications. Increasingly,
definitions of quality include an added emphasis on meeting the customers needs. As you
can see in Table M1.1, the first and second ones are similar to our definition.
Total quality management (TQM) refers to a quality emphasis that encompasses the
entire organization, from supplier to customer. TQM emphasizes a commitment by man-
HISTORY
M1-3
agement to have a companywide drive toward excellence in all aspects of the products
and services that are important to the customer. Meeting the customers expectations requires an emphasis on TQM if the firm is to compete as a leader in world markets.
M1-4
Module 1
FIGURE M1.1
Patterns to Look for on Control
Charts
(Source: Bertrand L. Hansen,
Quality Control: Theory and
Applications, 1963, renewed
1991, p. 65. Reprinted by
permission of Prentice Hall,
Upper Saddle River, NJ.)
Target
Target
Target
Erratic behavior.
Investigate.
Building Control Charts When building control charts, averages of small samples
(often of five items or parts) are used, as opposed to data on individual parts. Individual
pieces tend to be too erratic to make trends quickly visible. The purpose of control charts
is to help distinguish between natural variations and variations due to assignable
causes.
IN ACTION
Du Pont now saves more than 15 million pounds of plastics annually by recycling them into products rather than
dumping them into landfills. Through electronic purchasing
the firm has reduced wastepaper to a trickle and, by using new
packaging designs, has cut in-process material wastes by
nearly 40%.
By integrating SPC with environmental compliance activities, Du Pont has made major quality improvements that
far exceed regulatory guidelines, and at the same time the
company has realized huge cost savings.
Sources: Automotive Industries (June 1996): 93; and E. E. Dwinells and J. P.
Sheffer, APICSThe Performance Advantage (March 1992): 3031.
Natural Variations Natural variations affect almost every production process and are to
be expected. Natural variations are the many sources of variation within a process that is
in statistical control. They behave like a constant system of chance causes. Although individual measured values are all different, as a group they form a pattern that can be described as a distribution. When these distributions are normal, they are characterized by
two parameters. These parameters are
M1-5
1. Mean, (the measure of central tendency, in this case, the average value)
2. Standard deviation, (variation, the amount by which the smaller values differ from
the larger ones)
As long as the distribution (output precision) remains within specified limits, the process
is said to be in control, and the modest variations are tolerated.
Assignable Variations When a process is not in control, we must detect and eliminate
special (assignable) causes of variation. Then its performance is predictable, and its ability to meet customer expectations can be assessed. The ability of a process to operate
within statistical control is determined by the total variation that comes from natural
causesthe minimum variation that can be achieved after all assignable causes have been
eliminated. The objective of a process control system, then, is to provide a statistical signal when assignable causes of variation are present. Such a signal can quicken appropriate action to eliminate assignable causes.
Assignable variation in a process can be traced to a specific reason. Factors such as
machine wear, misadjusted equipment, fatigued or untrained workers, or new batches of
raw material are all potential sources of assignable variations. Control charts such as those
illustrated in Figure M1.1 help the manager pinpont where a problem may lie.
Assignable variations in a
process can be traced to a
specific problem.
Control charts for the mean, x, and the range, R, are used to monitor processes that are
measured in continuous units. The x- (x-bar) chart tells us whether changes have occurred
in the central tendency of a process. This might be due to such factors as tool wear, a
gradual increase in temperature, a different method used on the second shift, or new and
stronger materials. The R-chart values indicate that a gain or loss in uniformity has occurred. Such a change might be due to worn bearings, a loose tool part, an erratic flow of
lubricants to a machine, or to sloppiness on the part of a machine operator. The two types
of charts go hand in hand when monitoring variables.
and
sx
sx
n
M1-6
Module 1
FIGURE M1.2
Population and Sampling
Distributions
Normal
Beta
(mean)
x S.D.
Uniform
(mean)
x S.D.
(mean)
x S.D.
99.7% of all x
fall within 3x
95.5% of all x fall within 2x
3x
2x
1x
X
(mean)
Standard error x
1x
2x
3x
x
n
Figure M1.2 shows three possible population distributions, each with its own mean,
, and standard deviation, x. If a series of random samples (x1, x2, x3, x4, and so on) each
of size n is drawn from any one of these, the resulting distribution of xis will appear as in
the bottom graph of that figure. Because this is a normal distribution (as discussed in
Chapter 2), we can state that
1. 99.7% of the time, the sample averages will fall within 3sx if the process has only
random variations.
2. 95.5% of the time, the sample averages will fall within 2sx if the process has only
random variations.
If a point on the control chart falls outside the 3sx control limits, we are 99.7% sure
that the process has changed. This is the theory behind control charts.
(M1-1)
(M1-2)
M1-7
where
x mean of the sample means
z number of normal standard deviations (2 for 95.5% confidence, 3 for 99.7%)
s
sx standard deviation of the sample means x
n
Box-Filling Example Let us say that a large production lot of boxes of cornflakes is sampled every hour. To set control limits that include 99.7% of the sample means, 36 boxes
are randomly selected and weighed. The standard deviation of the overall population of
boxes is estimated, through analysis of old records, to be 2 ounces. The average mean of
all samples taken is 16 ounces. We therefore have x 16 ounces, x 2 ounces, n 36,
and z 3. The control limits are
2
36
16 1 17 ounces
2
16 3
16 1 15 ounces
36
UCL x x zsx 16 3
LCL x x zsx
Developing
a Model
AVX-Kyocera, a Japanese-owned maker of electronic chip components located in Raleigh, North Carolina, needed to improve the quality of its products and services to achieve total customer satisfaction.
Statistical process control models such as x- and R-charts were chosen as appropriate tools.
Acquiring
Input Data
Employees are empowered to collect their own data. For example, a casting machine operator measures
the thickness of periodic samples that he takes from his process.
Developing
a Solution
Employees plot data observations to generate SPC charts that track trends, comparing results with
process limits and final customer specifications.
Testing the
Solution
Samples at each machine are evaluated to ensure that the processes are indeed capable of achieving the
desired results. Quality control inspectors are transferred to manufacturing duties as all plant personnel
become trained in statistical methodology.
Analyzing
the Results
Implementing
the Results
Results of SPC are analyzed by individual operators to see if trends are present in their processes. Quality trend boards are posted in the building to display not only SPC charts, but also procedures, process
document change approvals, and the names of all certified operators. Work teams are in charge of analysis of clusters of machines.
The firm has implemented a policy of zero defectives at a very low tolerance for variable data and nearly
zero defects for parts per million for attribute data.
Source: Basile A. Denisson. War with Defects and Peace with Quality, Quality Progress (September 1993): 97101.
M1-8
Module 1
If the process standard deviation is not available or is difficult to compute, which is usually the case, these equations become impractical. In practice, the calculation of control limits
is based on the average range rather than on standard deviations. We may use the equations
Control chart limits can be
found using the range rather
than the standard deviation.
UCL x x A2R
(M1-3)
LCL x x A2R
(M1-4)
where
R average of the samples
A2 value found in Table M1.2 (which assumes that Z 3)
x mean of the sample means
Here is an example. Super Cola bottles soft drinks labeled net weight 16 ounces. An
overall process average of 16.01 ounces has been found by taking several batches of samples,
where each sample contained five bottles. The average range of the process is 0.25 ounce.
We want to determine the upper and lower control limits for averages for this process.
Looking in Table M1.2 for a sample size of 5 in the mean factor A2 column, we find
the number 0.577. Thus the upper and lower control chart limits are
UCLx x A2R
16.01 (0.577)(0.25)
16.01 0.144
16.154
LCLx x A2R
16.01 0.144
15.866
The upper control limit is 16.154, and the lower control limit is 15.866.
We just determined the upper and lower control limits for the process average. In addition
to being concerned with the process average, managers are interested in the dispersion or
variability. Even though the process average is under control, the variability of the
process may not be. For example, something may have worked itself loose in a piece of
equipment. As a result, the average of the samples may remain the same, but the variation
within the samples could be entirely too large. For this reason it is very common to find a
control chart for ranges in order to monitor the process variability. The theory behind the
control charts for ranges is the same for the process average. Limits are established that
contain 3 standard deviations of the distribution for the average range R. With a few
simplifying assumptions, we can set the upper and lower control limits for ranges:
UCLR D4R
LCLR D3R
where
UCLR upper control chart limit for the range
LCLR lower control chart limit for the range
D4 and D3 values from Table M1.2
(M1-5)
(M1-6)
M1-9
MEAN FACTOR, A2
UPPER RANGE, D4
LOWER RANGE, D3
1.880
3.268
1.023
2.574
0.729
2.282
0.577
2.114
0.483
2.004
0.419
1.924
0.076
0.373
1.864
0.136
0.337
1.816
0.184
10
0.308
1.777
0.223
12
0.266
1.716
0.284
14
0.235
1.671
0.329
16
0.212
1.636
0.364
18
0.194
1.608
0.392
20
0.180
1.586
0.414
25
0.153
1.541
0.459
Source: Reprinted by permission of the American Society for Testing and Materials, copyright. Taken from Special Technical
Publication 15-C, Quality Control of Materials, pp. 63 and 72, 1951.
Control charts for x and R do not apply when we are sampling attributes, which are typically classified as defective or nondefective. Measuring defectives involves counting them
(for example, number of bad lightbulbs in a given lot, or number of letters or data entry
M1-10
Module 1
records typed with errors); whereas variables are usually measured for length or weight.
There are two kinds of attribute control charts: (1) those that measure the percent defective in a sample, called p-charts, and (2) those that count the number of defects, called
c-charts.
p-Charts
p-chart limits are based on
the binomial distribution
and are easy to compute.
p-charts are the principal means of controlling attributes. Although attributes that are either good or bad follow the binomial distribution, the normal distribution can be used to
calculate p-chart limits when sample sizes are large. The procedure resembles the x-chart
approach, which was also based on the central limit theorem.
The formulas for p-chart upper and lower control limits follow:
UCLp p zsp
(M1-7)
LCLp p zsp
(M1-8)
where
p mean fraction defective in the sample
z number of standard deviates (z = 2 for 95.5% limits; z = 3 for 99.7% limits)
sp standard deviation of the sampling distribution
p(1 p)
n
(M1-9)
SAMPLE
NUMBER
NUMBER
OF ERRORS
FRACTION
DEFECTIVE
SAMPLE
NUMBER
NUMBER
OF ERRORS
0.06
11
0.06
0.05
12
0.01
0.00
13
0.08
0.01
14
0.07
0.04
15
0.05
0.02
16
0.04
0.05
17
11
0.11
0.03
18
0.03
0.03
19
0.00
10
0.02
20
04
0.04
M1-11
FRACTION
DEFECTIVE
80
We want to set control limits that include 99.7% of the random variation in the entry
process when it is in control. Thus, z 3.
p
sp
c-Charts
In the ARCO example above, we counted the number of defective database records entered. A defective record was one that was not exactly correct. A bad record may contain
more than one defect, however. We use c-charts to control the number of defects per unit
of output (or per insurance record in the case above).
M1-12
Module 1
Fraction defective
FIGURE M1.3
p-Chart for Data Entry for
ARCO
0.12
0.11
0.10
0.09
0.08
0.07
UCLp 0.10
0.06
0.05
0.04
0.03
0.02
p 0.04
0.01
LCLp 0.00
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Sample number
Control charts for defects are helpful for monitoring processes in which a large number of potential errors can occur but the actual number that do occur is relatively small.
Defects may be mistyped words in a newspaper, blemishes on a table, or missing pickles
on a fast-food hamburger.
The Poisson probability distribution, which has a variance equal to its mean, is the
basis for c-charts. Since c is the mean number of defects per unit, the standard deviation is
equal to c. To compute 99.7% control limits for c, we use the formula
c 3c
PROGRAM M1.1A
Excel QMs p-Chart Program
Applied to the ARCO Data,
Showing Input Data and
Formulas
(M1-10)
Summary
M1-13
PROGRAM M1.1B
Output from Excel QMs
p-Chart Analysis of the
ARCO Data
Here is an example.
Red Top Cab Company c-Chart Example Red Top Cab Company receives several
complaints per day about the behavior of its drivers. Over a nine-day period (where days
are the units of measure), the owner received the following numbers of calls from irate
passengers: 3, 0, 8, 9, 6, 7, 4, 9, 8, for a total of 54 complaints.
To compute 99.7% control limits, we take
c
54
6 complaints per day
9
Thus,
LCLc c 3c 6 36 6 3(2.45) 0
Summary
To the manager of a firm producing goods or services, quality is the degree to which the
product meets specifications. Quality control has become one of the most important precepts of business.
The expression quality cannot be inspected into a product is a central theme of organizations today. More and more world-class companies are following the ideas of total quality
management (TQM), which emphasizes the entire organization, from supplier to customer.
M1-14
Module 1
Statistical aspects of quality control date to the 1920s but are of special interest in our
global marketplaces of this new century. Statistical process control (SPC) tools described
in this chapter include the x- and R-charts for variable sampling and the p- and c-charts
for attribute sampling.
Glossary
Quality. The degree to which a product or service meets the specifications set for it.
Total Quality Management (TQM). An emphasis on quality that encompasses the entire organization.
Control Chart. A graphic presentation of process data over time.
Natural Variations. Variabilities that affect almost every production process to some degree and
are to be expected; also known as common causes.
Assignable Variation. Variation in the production process that can be traced to specific causes.
x-Chart. A quality control chart for variables that indicates when changes occur in the central tendency of a production process.
R-Chart. A process control chart that tracks the range within a sample; indicates that a gain or
loss of uniformity has occurred in a production process.
Central Limit Theorem. The theoretical foundation for x-charts. It states that regardless of the
distribution of the population of all parts or services, the distribution of xs will tend to follow a
normal curve as the sample size grows.
p-Chart. A quality control chart that is used to control attributes.
c-Chart. A quality control chart that is used to control the number of defects per unit of output.
Key Equations
(M1-1) Upper control limit (UCL) x Zsx
The upper limit for an x-chart using standard deviations.
(M1-2) Lower control limit (LCL) x Zsx
The lower control limit for an x-chart using standard deviations.
(M1-3) UCLx x A2R
The upper control limit for an x-chart using tabled values and ranges.
(M1-4) LCLx x A2R
The lower control limit for an x-chart using tabled values and ranges.
(M1-5) UCLR D4R
Upper control limit for a range chart.
(M1-6) LCLR D3R
Lower control limit for a range chart.
(M1-7) UCLp p zsp
Upper control unit for a p-chart.
(M1-8) LCLp p zsp
Lower control limit for a p-chart.
p(1 p)
(M1-9) sp
n
Solved Problems
Solved Problems
Solved Problem M1-1
The manufacturer of precision parts for drill presses produces round shafts for use in the construction of drill presses. The average diameter of a shaft is 0.56 inch. The inspection samples
contain six shafts each. The average range of these samples is 0.006 inch. Determine the upper
and lower control chart limits.
Solution
The mean factor A2 from Table M1.2, where the sample size is 6, is seen to be 0.483. With this
factor, you can obtain the upper and lower control limits:
UCLx 0.56 (0.483)(0.006)
0.5571
SAMPLE
SAMPLE
RANGE
SAMPLE
AVERAGE
SAMPLE
SAMPLE
RANGE
SAMPLE
AVERAGE
0.41
4.00
0.56
4.17
0.55
4.16
0.62
3.93
0.44
3.99
0.54
3.98
0.48
4.00
0.44
4.01
M1-15
M1-16
Module 1
Solution
UCLp p 3
p(1 p)
(0.05)(1 0.05)
0.05 3
n
100
LCLp p 3
p(1 p)
0.05 3(0.0218)
n
Self-Test
M1-17
SELF-TEST
Before taking the self-test, refer back to the learning objectives at the beginning of the
module, the notes in the margins, and the glossary at the end of the module.
Use the key at the back of the book to correct your answers.
Restudy pages that correspond to any questions that you answered incorrectly or material
you feel uncertain about.
7.
8.
9.
10.
11.
M1-18
Module 1
M1-2
M1-3
M1-4
M1-5
Explain why a process can be out of control even though all the samples fall within
the upper and lower control limits.
Problems*
M1-6 Shader Storage Technologies produces refrigeration units for food producers and retail food establishments. The overall average temperature that these units maintain is
46 Fahrenheit. The average range is 2 Fahrenheit. Samples of 6 are taken to monitor
the process. Determine the upper and lower control chart limits for averages and
ranges for these refrigeration units.
M1-7
M1-8
M1-9
When set at the standard position, Autopitch can throw hard balls toward a batter at
an average speed of 60 mph. Autopitch devices are made for both major- and minorleague teams to help them improve their batting averages. Autopitch executives take
samples of 10 Autopitch devices at a time to monitor these devices and to maintain
the highest quality. The average range is 3 mph. Using control-chart techniques, determine control-chart limits for averages and ranges for Autopitch.
Zipper Products, Inc., produces granola cereal, granola bars, and other natural food
products. Its natural granola cereal is sampled to ensure proper weight. Each sample
contains eight boxes of cereal. The overall average for the samples is 17 ounces. The
range is only 0.5 ounce. Determine the upper and lower control-chart limits for averages for the boxes of cereal.
Small boxes of NutraFlakes cereal are labeled net weight 10 ounces. Each hour,
random samples of size n 4 boxes are weighed to check process control. Five hours
of observations yielded the following:
WEIGHT
Time
* Note:
Box 1
Box 2
Box 3
Box 4
9 A.M.
9.8
10.4
9.9
10.3
10 A.M.
10.1
10.2
9.9
9.8
11 A.M.
9.9
10.5
10.3
10.1
Noon
9.7
9.8
10.3
10.2
1 A.M.
9.7
10.1
9.9
9.9
Using these data, construct limits for x- and R-charts. Is the process in control? What
other steps should the QC department follow at this point?
M1-10 Sampling four pieces of precision-cut wire (to be used in computer assembly) every
hour for the past 24 hours has produced the following results:
HOUR
HOUR
3.25
0.71
13
3.11
0.85
3.10
1.18
14
2.83
1.31
3.22
1.43
15
3.12
1.06
3.39
1.26
16
2.84
0.50
3.07
1.17
17
2.86
1.43
2.86
0.32
18
2.74
1.29
3.05
0.53
19
3.41
1.61
2.65
1.13
20
2.89
1.09
3.02
0.71
21
2.65
1.08
10
2.85
1.33
22
3.28
0.46
11
2.83
1.17
23
2.94
1.58
12
2.97
0.40
24
2.64
0.97
Develop appropriate control limits and determine whether there is any cause for concern in the cutting process.
M1-11 Due to the poor quality of various semiconductor products used in their manufacturing process, Microlaboratories has decided to develop a quality control program. Because the semiconductor parts they get from suppliers are either good or defective,
Milton Fisher has decided to develop control charts for attributes. The total number of
semiconductors in every sample is 200. Furthermore, Milton would like to determine
the upper control chart limit and the lower control chart limit for various values of the
fraction defective (p) in the sample taken. To allow more flexibility, he has decided to
develop a table that lists value for p, UCL, and LCL. The values for p should range
from 0.01 to 0.1, incrementing by 0.01 each time. What are the UCLs and the LCLs
for 99.7% confidence?
M1-12 For the past two months, Suzan Shader has been concerned about machine number 5
at the West Factory. To make sure that the machine is operating correctly, samples
are taken, and the average and range for each sample is computed. Each sample consists of 12 items produced from the machine. Recently, 12 samples were taken, and
for each, the sample range and average were computed. The sample range and sample
average were 1.1 and 46 for the first sample, 1.31 and 45 for the second sample, 0.91
and 46 for the third sample, and 1.1 and 47 for the fourth sample. After the fourth
sample, the sample averages increased. For the fifth sample, the range was 1.21 and
the average was 48; for sample number 6 it was 0.82 and 47; for sample number 7, it
was 0.86 and 50; and for the eighth sample, it was 1.11 and 49. After the eighth sample, the sample average continued to increase, never getting below 50. For sample
number 9, the range and average were 1.12 and 51; for sample number 10, they were
M1-19
M1-20
Module 1
0.99 and 52; for sample number 11, they were 0.86 and 50; and for sample number
12, they were 1.2 and 52.
Although Suzans boss wasnt overly concerned about the process, Suzan was.
During installation, the supplier set a value of 47 for the process average with an average range of 1.0. It was Suzans feeling that something was definitely wrong with
machine number 5. Do you agree?
M1-13 Kitty Products caters to the growing market for cat supplies, with a full line of products, ranging from litter to toys to flea powder. One of its newer products, a tube of
fluid that prevents hair balls in long-haired cats, is produced by an automated machine
that is set to fill each tube with 63.5 grams of paste.
To keep this filling process under control, four tubes are pulled randomly from the
assembly line every 4 hours. After several days, the data shown in the following table
resulted. Set control limits for this process and graph the sample data for both the xand R-charts.
Sample
No.
Sample
No.
Sample
No.
63.5
2.0
10
63.5
1.3
18
63.6
1.8
63.6
1.0
11
63.3
1.8
19
63.8
1.3
63.7
1.7
12
63.2
1.0
20
63.5
1.6
63.9
0.9
13
63.6
1.8
21
63.9
1.0
63.4
1.2
14
63.3
1.5
22
63.2
1.8
63.0
1.6
15
63.4
1.7
23
63.3
1.7
63.2
1.8
16
63.4
1.4
24
64.0
2.0
63.3
1.3
17
63.5
1.1
25
63.4
1.5
63.7
1.6
M1-14 The smallest defect in a computer chip will render the entire chip worthless. Therefore, tight quality control measures must be established to monitor the chips. In the
past, the percentage defective for these chips for a California-based company has
been 1.1%. The sample size is 1,000. Determine the upper and lower control-chart
limits for these computer chips. Use z 3.
M1-15 Barbara Schwartzs Office Supply Company manufactures paper clips and other
office products. Although inexpensive, paper clips have provided Barbara with a
high margin of profitability. The percentage defective for paper clips produced by
Office Supply Company has been averaging 2.5%. Samples of 200 paper clips are
taken. Establish the upper and lower control-chart limits for this process at 99.7%
confidence.
M1-16 Daily samples of 100 power drills are removed from Drill Masters assembly line and
inspected for defects. Over the past 21 days, the following information has been gathered. Develop a 3 standard deviation (99.7% confidence) p-chart and graph the samples. Is the process in control?
Case Study
DAY
NUMBER OF
DEFECTIVE DRILLS
DAY
NUMBER OF
DEFECTIVE DRILLS
12
13
14
15
16
17
18
19
20
10
21
11
M1-21
M1-17 A random sample of 100 Modern Art dining room tables that came off the firms assembly line is examined. Careful inspection reveals a total of 2,000 blemishes. What
are the 99.7% upper and lower control limits for the number of blemishes? If one
table had 42 blemishes, should any special action be taken?
Case Study
Bayfield Mud Company
In November 1998, John Wells, a customer service representative of Bayfield Mud Company, was summoned to the Houston, Texas, warehouse of Wet-Land Drilling, Inc., to inspect
three boxcars of mud treating agents that Bayfield Mud Company had shipped to the Houston firm. (Bayfields corporate
offices and its largest plant are located in Orange, Texas,
which is just west of the LouisianaTexas border.) Wet-Land
Drilling had filed a complaint that the 50-pound bags of treating agents that it had just received from Bayfield were shortweighted by approximately 5%.
The light-weight bags were initially detected by one of
Wet-Lands receiving clerks, who noticed that the railroad
scale tickets indicated that the net weights were significantly
less on all three of the boxcars than those of identical shipments received on October 25, 1998. Bayfields traffic department was called to determine whether lighter-weight dunnage
or pallets were used on the shipments. (This might explain the
lighter net weights.) Bayfield indicated, however, that no
changes had been made in the loading or palletizing procedures. Hence, Wet-Land randomly checked 50 of the bags and
discovered that the average net weight was 47.51 pounds. They
noted from past shipments that the bag net weights averaged
exactly 50.0 pounds, with an acceptable standard deviation of
1.2 pounds. Consequently, they concluded that the sample indicated a significant short-weight. (The reader may wish to verify
this conclusion.) Bayfield was then contacted, and Wells was
sent to investigate the complaint. Upon arrival, Wells verified
the complaint and issued a 5% credit to Wet-Land.
Wet-Lands management, however, was not completely
satisfied with only the issuance of credit for the short shipment.
The charts followed by their mud engineers on the drilling platforms were based on 50-pound bags of treating agents. Lighterweight bags might result in poor chemical control during the
drilling operation and might adversely affect drilling efficiency.
(Mud treating agents are used to control the pH and other
chemical properties of the cone during drilling operation.) This
could cause severe economic consequences because of the extremely high cost of oil and natural gas well drilling operations.
Consequently, special use instructions had to accompany the
delivery of these shipments to the drilling platforms. Moreover,
the light-weight shipments had to be isolated in Wet-Lands
warehouse, causing extra handling and poor space utilization.
Hence, Wells was informed that Wet-Land Drilling might seek
a new supplier of mud treating agents if, in the future, it received bags that deviated significantly from 50 pounds.
The quality control department at Bayfield suspected that
the light-weight bags may have resulted from growing pains
M1-22
Module 1
TIME
AVERAGE
WEIGHT
(POUNDS)
TIME
AVERAGE
WEIGHT
(POUNDS)
Smallest
Largest
Smallest
Largest
6:00 A.M.
49.6
48.7
50.7
6:00 A.M.
46.8
41.0
51.2
7:00
50.2
8:00
50.6
49.1
51.2
7:00
50.0
46.2
51.7
49.6
51.4
8:00
47.4
44.0
48.7
9:00
50.8
50.2
51.8
9:00
47.0
44.2
48.9
10:00
49.9
49.2
52.3
10:00
47.2
46.6
50.2
11:00
50.3
48.6
51.7
11:00
48.6
47.0
50.0
12 noon
48.6
46.2
50.4
12 midnight
49.8
48.2
50.4
1:00 P.M.
49.0
46.4
50.0
1:00 A.M.
49.6
48.4
51.7
2:00
49.0
46.0
50.6
2:00
50.0
49.0
52.2
3:00
49.8
48.2
50.8
3:00
50.0
49.2
50.0
4:00
50.3
49.2
52.7
4:00
47.2
46.3
50.5
5:00
51.4
50.0
55.3
5:00
47.0
44.1
49.7
6:00
51.6
49.2
54.7
6:00
48.4
45.0
49.0
7:00
51.8
50.0
55.6
7:00
48.8
44.8
49.7
8:00
51.0
48.6
53.2
8:00
49.6
48.0
51.8
9:00
50.5
49.4
52.4
9:00
50.0
48.1
52.7
10:00
49.2
46.1
50.7
10:00
51.0
48.1
55.2
11:00
49.0
46.3
50.8
11:00
50.4
49.5
54.1
12 midnight
48.4
45.4
50.2
12 noon
50.0
48.7
50.9
1:00 A.M.
47.6
44.3
49.7
1:00 P.M.
48.9
47.6
51.2
2:00
47.4
44.1
49.6
2:00
49.8
48.4
51.0
3:00
48.2
45.2
49.0
3:00
49.8
48.8
50.8
4:00
48.0
45.5
49.1
4:00
50.0
49.1
50.6
5:00
48.4
47.1
49.6
5:00
47.8
45.2
51.2
6:00
48.6
47.4
52.0
6:00
46.4
44.0
49.7
7:00
50.0
49.2
52.2
7:00
46.4
44.4
50.0
8:00
49.8
49.0
52.4
8:00
47.2
46.6
48.9
9:00
50.3
49.4
51.7
9:00
48.4
47.2
49.5
10:00
50.2
49.6
51.8
10:00
49.2
48.1
50.7
11:00
50.0
49.0
52.3
11:00
48.4
47.0
50.8
12 noon
50.0
48.8
52.4
12 midnight
47.2
46.4
49.2
1:00 A.M.
50.1
49.4
53.6
1:00 A.M.
47.4
46.8
49.0
2:00
49.7
48.6
51.0
2:00
48.8
47.2
51.4
3:00
48.4
47.2
51.7
3:00
49.6
49.0
50.6
4:00
47.2
45.3
50.9
4:00
51.0
50.5
51.5
5:00
46.8
44.1
49.0
5:00
50.5
50.0
51.9
RANGE
RANGE
M1-23
Case Study
Case Study
Morristown Daily Tribune
In July 1998, the Morristown Daily Tribune published its first
newspaper in direct competition with two other newspapers,
the Morristown Daily Ledger and the Clarion Herald, a
weekly publication. Presently, the Ledger is the most widely
read newspaper in the area, with a total circulation of 38,500.
The Tribune, however, has made significant inroads into the
readership market since its inception. Total circulation of the
Tribune now exceeds 27,000.
Rita Bornstein, editor of the Tribune, attributes the success of the newspaper to the accuracy of its contents, a strong
editorial section, and the proper blending of local, regional,
national, and international news items. In addition, the paper
has been successful in getting the accounts of several major retailers who advertise extensively in the display section. Finally, experienced reporters, photographers, copy writers,
PARAGRAPHS
WITH ERRORS
SAMPLE IN THE SAMPLE
FRACTION
OF PARAGRAPHS
WITH ERRORS
(PER 100)
SAMPLE
PARAGRAPHS
WITH ERRORS
IN THE SAMPLE
FRACTION
OF PARAGRAPHS
WITH ERRORS
(PER 100)
0.02
16
0.02
0.04
17
0.03
10
0.10
18
0.07
0.04
19
0.03
0.01
20
0.02
0.01
21
0.03
M1-24
Module 1
PARAGRAPHS
WITH ERRORS
IN THE SAMPLE
FRACTION
OF PARAGRAPHS
WITH ERRORS
(PER 100)
SAMPLE
PARAGRAPHS
WITH ERRORS
IN THE SAMPLE
FRACTION
OF PARAGRAPHS
WITH ERRORS
(PER 100)
13
0.13
22
0.07
0.09
23
0.04
11
0.11
24
0.03
10
0.00
25
0.02
11
0.03
26
0.02
12
0.04
27
0.00
13
0.02
28
0.01
14
0.02
29
0.03
15
0.08
30
0.04
SAMPLE
Discussion Questions
1. Plot the overall fraction of errors ( p) and the upper and
lower control limits on a control chart using a 95% confidence level.
2. Assume that the industry upper and lower control limits
are 0.1000 and 0.0400, respectively. Plot them on the
control chart.
Bibliography
Berry, L. L., A. Parasuraman, and V. A. Zeithaml.
Improving Service Quality in America:
Lessons Learned, The Academy of
Management Executive 8, 2 (May 1994): 3252.
M1-25
M O D U L E
Dynamic
Programming
LEARNING OBJECTIVES
After completing this module, students will be able to:
1.
2.
3.
4.
5.
MODULE OUTLINE
M2.1
M2.2
M2.3
M2.4
M2.5
Introduction
Shortest-Route Problem Solved by Dynamic Programming
Dynamic Programming Terminology
Dynamic Programming Notation
Knapsack Problem
Summary Glossary Key Equations Solved
Problem Self-Test Discussion Questions and
Problems Case Study: United Trucking Internet Case
Study Bibliography
M2-1
M2-2
Module 2
DYNAMIC PROGRAMMING
M2.1 INTRODUCTION
Dynamic programming
breaks a difficult problem
into subproblems.
Dynamic programming is a quantitative analysis technique that has been applied to large,
complex problems that have sequences of decisions to be made. Dynamic programming divides problems into a number of decision stages; the outcome of a decision at one stage affects the decision at each of the next stages. The technique is useful in a large number of
multiperiod business problems, such as smoothing production employment, allocating capital
funds, allocating salespeople to marketing areas, and evaluating investment opportunities.
Dynamic programming differs from linear programming in two ways. First, there is
no algorithm (like the simplex method) that can be programmed to solve all problems.
Dynamic programming is, instead, a technique that allows us to break up difficult problems into a sequence of easier subproblems, which are then evaluated by stages. Second,
linear programming is a method that gives single-stage (one time period) solutions. Dynamic programming has the power to determine the optimal solution over a one-year time
horizon by breaking the problem into 12 smaller one-month time horizon problems and to
solve each of these optimally. Hence it uses a multistage approach.
Solving problems with dynamic programming involves four steps:
Four Steps of Dynamic Programming
1. Divide the original problem into subproblems called stages.
2. Solve the last stage of the problem for all possible conditions or states.
3. Working backward from the last stage, solve each intermediate stage. This is done by
determining optimal policies from that stage to the end of the problem (last stage).
4. Obtain the optimal solution for the original problem by solving all stages sequentially.
In this module we show you how to solve two types of dynamic programming problems: network and nonnetwork. The shortest-route problem is a network problem that can
be solved by dynamic programming. The knapsack problem is an example of a nonnetwork problem that can be solved using dynamic programming.
PROGRAMMING
George Yates is about to make a trip from Rice, Georgia (1) to Dixieville, Georgia (7).
George would like to find the shortest route. Unfortunately, there are a number of small
towns between Rice and Dixieville. His road map is shown in Figure M2.1.
The circles on the map, called nodes, represent cities such as Rice, Dixieville, Brown,
and so on. The arrows, called arcs, represent highways between the cities. The distance in
miles is indicated along each arc. This problem can, of course, be solved by inspection.
But seeing how dynamic programming can be used on this simple problem will teach you
how to solve larger and more complex problems.
Step 1: The first step is to divide the problem into subproblems or stages. Figure M2.2 reveals the stages of this problem. In dynamic programming, we usually start with the last
part of the problem, stage 1, and work backward to the beginning of the problem or network, which is stage 3 in this problem. Table M2.1 summarizes the arcs and arc distances
for each stage.
FIGURE M2.1
Highway Map between Rice and Dixieville
Lakecity
Athens
10 Miles
5
14
les
ile
s
i
2M
Mi
7 Dixieville
5 Miles
Brown
Mi
les
les
6M
ile
ile
2M
ile
s
Rice
10 Miles
2
Hope
6
Georgetown
FIGURE M2.2
Three Stages for the George
Yates Problem
A Node
A Branch
5 14
10
5
2
12
6
7
2
10
2
Stage 3
Stage 2
TABLE M2.1
Stage 1
STAGE
ARC
ARC DISTANCE
57
14
67
45
10
35
12
36
25
26
10
14
13
12
M2-3
M2-4
Module 2
DYNAMIC PROGRAMMING
FIGURE M2.3
Solution for the One-Stage
Problem
5 14
4
1
Minimum Distance
to Node 7 from
Node 5
14
10
12
3 6
5
2
7
2
10
2
2
Minimum Distance
to Node 7 from
Node 6
Step 2: We next solve stage 1, the last part of the network. Usually, this is trivial. We find
the shortest path to the end of the network, node 7 in this problem. At stage 1, the shortest
paths from node 5, and node 6 to node 7 are the only paths. You may also note in Figure
M2.3 that the minimum distances are enclosed in boxes by the entering nodes to stage 1,
node 5 and node 6. The objective is to find the shortest distance to node 7. The following
table summarizes this procedure for stage 1. As mentioned previously, the shortest distance is the only distance at stage 1.
STAGE 1
BEGINNING
NODE
SHORTEST DISTANCE
TO NODE 7
ARCS ALONG
THIS PATH
14
57
67
Step 3: Moving backward, we now solve for stages 2 and 3. At stage 2 we will use Figure
M2.4.
If we are at node 4, the shortest and only route to node 7 is arcs 45 and 57. At node
3, the shortest route is arcs 36 and 67 with a total minimum distance of 8 miles. If we
are at node 2, the shortest route is arcs 26 and 67 with a minimum total distance of 12
miles. This information is summarized in the stage 2 table:
STAGE 2
BEGINNING
NODE
SHORTEST DISTANCE
TO NODE 7
ARCS ALONG
THIS PATH
24
45
57
36
67
12
26
67
Minimum Distance
to Node 7 from
Node 4
M2-5
FIGURE M2.4
Solution for the Two-Stage
Problem
24
4
10
Minimum Distance
to Node 7 from
Node 3
14
8
12
5
2
14
6
2
4
10
2
2
Minimum Distance
to Node 7 from
Node 2
12
The solution to stage 3 can be completed using the accompanying table and the network in Figure M2.5.
STAGE 3
BEGINNING
NODE
SHORTEST DISTANCE
TO NODE 7
ARCS ALONG
THIS PATH
13
13
36
67
Step 4: To obtain the optimal solution at any stage, all we consider are the arcs to the
next stage and the optimal solution at the next stage. For stage 3, we only have to consider
the three arcs to stage 2 (12, 13, and 14) and the optimal policies at stage 2, given in a
previous table. This is how we arrived at the preceding solution. When the procedure is
understood, we can perform all the calculations on one network. You may want to study
the relationship between the networks and the tables because more complex problems are
usually solved by using tables only.
Regardless of the type or size of a dynamic programming problem, there are some important
terms and concepts that are inherent in every problem. Some of the more important follow:
1. Stage: a period or a logical subproblem.
2. State variables: possible beginning situations or conditions of a stage. These have
also been called the input variables.
M2-6
Module 2
DYNAMIC PROGRAMMING
FIGURE M2.5
Solution for the Three-Stage
Problem
24
Minimum Distance
to Node 7 from
Node 1
13
10
14
5 14
4
1
5
2
12
3 6
7
2
10
2
2
12
IN ACTION
Managing a nursery that produces ornamental plants is difficult. In most cases, ornamental plants increase in value with
increased growth. This value-added growth makes it difficult
to determine when to harvest the plants and place them on the
market. When plants are marketed earlier, revenues are generated earlier and the costs associated with plant growth are
minimized. On the other hand, delaying the harvesting of the
ornamental plants usually results in higher prices. But are the
additional months of growth and costs worth the delay?
In this case, dynamic programming was used to determine the optimal growth stages for ornamental plants. Each
stage was associated with a possible growth level. The state
variables included acres of production of ornamental plants
and carryover plants from previous growing seasons. The ob-
jective of the dynamic programming problem was to maximize the after-tax cash flow. The taxes included self-employment, federal income, earned income credit, and state income
taxes.
The solution was to produce one- and three-gallon containers of ornamental plants. The one-gallon containers are
sold in the fall and carried over for spring sales. Any one-gallon containers not sold in the spring are combined into threegallon container products for sale during the next season.
Using dynamic programming helps to determine when to harvest to increase after-tax cash flow.
Source: Stokes, Jeffery et al. Optimal Marketing of Nursery Crops From
Container-Based Production Systems, American Journal of Agricultural
Economics (February 1997): 235.
M2-7
This relationship shows how we were able to go from one stage to the next in solving
for the optimal solution to the shortest-route problem. In more complex problems, we can
use symbols to show the relationship between stages.
State variables, decision variables, the decision criterion, and the optimal policy can
be determined for any stage of a dynamic programming problem. This is done here for
stage 2 of the George Yates shortest-route problem.
1. State variables for stage 2 are the entering nodes, which are
(a) Node 2
(b) Node 3
(c) Node 4
2. Decision variables for stage 2 are the following arcs or routes:
(a) 45
(b) 35
(c) 36
(d) 25
(e) 26
3. The decision criterion is the minimization of the total distances traveled.
4. The optimal policy for any beginning condition is shown in Figure M2.6 and
following the table below.
GIVEN THIS ENTERING
CONDITION
26
36
45
Figure M2.6 may also be helpful in understanding some of the terminology used in
the discussion of dynamic programming.
M2-8
Module 2
DYNAMIC PROGRAMMING
FIGURE M2.6
Stage 2 from the ShortestRoute Problem
State
variables
are the
entering
nodes.
24
10
14
Decision
variables
are all
the arcs.
8
5
12
3 6
7
6
10
2
2
The optimal policy is the arc,
for any entering node, that
will minimize total distance
to the destination at this
stage.
12
Stage 2
lead to stage 1, is represented by d2. The possible arcs or decisions are 45, 35, 36, and
26. The output to stage 2 becomes the input to stage 1. The output from stage 2 is s1. The
possible outputs from stage 2 are the exiting nodes, nodes 5 and 6. Finally, each stage has
a return. For stage 2, the return is represented by r2. In our shortest-route problem, the return is the distance along the arcs in stage 2. These distances are 10 miles for arc 45, 12
FIGURE M2.7
Input, Decision, Output, and
Return for Stage 2 in George
Yatess Problem
Decision
d2
Stage 2
Input
s2
Output
s1
Return
r2
M2-9
miles for arc 35, 6 miles for arc 36, and 10 miles for arc 26. The same notation applies
for the other stages and can be used at any stage. In general, we will use the following notation for these important concepts:
sn input to stage n
(M2-1)
dn decision at stage n
(M2-2)
rn return at stage n
(M2-3)
You should also note that the input to one stage is also the output from another stage.
For example, the input to stage 2, s2, is also the output from stage 3 (see Figure M2.7).
This leads us to the following equation:
sn 1 output from stage n
(M2-4)
A transformation function
allows us to go from one
stage to another.
(M2-5)
The following general formula allows us to go from one stage to another using the transformation function:
sn1 tn (sn , dn)
(M2-6)
Although this equation may seem complex, it is really just a mathematical statement of
the fact that the output from a stage is a function of the input to the stage and any decisions made at that stage. In the George Yates shortest-route problem, the transformation
function consisted of a number of tables. These tables showed how we could progress
from one stage to another in order to solve the problem. For more complex problems, we
need to use dynamic programming notation instead of tables.
Another useful quantity is the total return at any stage. The total return allows us to
keep track of the total profit or costs at each stage as we solve the dynamic programming
problem. It can be given as follows:
fn total return at stage n
(M2-7)
The knapsack problem involves the maximization or the minimization of a value, such as
profits or costs. Like a linear programming problem, there are restrictions. Imagine a
knapsack or pouch that can only hold a certain weight or volume. We can place different
types of items in the knapsack. Our objective will be to place items in the knapsack to
maximize total value without breaking the knapsack because of too much weight or a similar restriction.
M2-10
Module 2
DYNAMIC PROGRAMMING
or both. Some scheduling problems are also knapsack problems. For example, we may
want to determine which jobs to complete in the next two weeks. The two-week period is
the knapsack, and we want to load it with jobs in such a way as to maximize profits or
minimize costs. The restriction is the number of days or hours during the two-week period.
Rob Roller owns and operates Rollers Air Transport Service, which ships cargo by plane
to most large cities in the United States and Canada. The remaining capacity for one of
the flights from Seattle to Vancouver is 10 tons. There are four different items that Rob
can ship between Seattle and Vancouver. Each item has a weight in tons, a net profit in
thousands of dollars, and a total number of that item that is available for shipping. This information is presented in Table M2.2.
Developing
a Model
The Southern Company, with service areas in Georgia, Alabama, Mississippi, and Florida, is a major
provider of electric service, with about 240 generating units. In recent years, fuel costs have increased
faster than other costs. Annual fuel costs are about $2.5 billion, representing about one-third of total expenses for the Southern Company. The problem for the Southern Company is to reduce total fuel costs.
To deal with this fuel cost problem, the company developed a state-of-the-art dynamic programming
model. The dynamic programming model is embedded in the Wescouger optimization program, which
is a computer program used to control electric generating units and reduce fuel costs through better utilization of existing equipment.
Acquiring
Input Data
Data were collected on past and projected electric usage. In addition, daily load/generation data were
analyzed. Load/generation charts were used to investigate the fuel requirements for coal, nuclear, hydroelectric, and gas/oil.
Developing
a Solution
The solution of the dynamic programming model provides both short-term planning guidelines and
long-term fuel usage for the various generating units. Optimal maintenance schedules for generating
units are obtained using Wescouger.
Testing the
Solution
Analyzing
the Results
Implementing
the Results
To test the accuracy of the Wescouger optimization program, Southern used a real-time economic dispatch program. The results were a very close match. In addition, the company put the solution through
an acid test, in which seasoned operators compared the results against their intuitive judgment. Again,
the results were consistent.
The results were analyzed in terms of their impact on the use of various fuels, the usage of various generating units, and maintenance schedules for generating units. Analyzing the results also revealed other
needs. This resulted in a full-color screen editing routine, auxiliary programs to automate data input, and
software to generate postanalysis reports.
The Southern Company implemented the dynamic programming solution. Over a seven-year period, the
results saved the company over $140 million.
Source: S. Erwin, et al. Using an Optimization Software to Lower Overall Electric Production Costs for Southern Company,
Interfaces 21, 1 (JanuaryFebruary 1991): 2741.
TABLE M2.2
TABLE M2.3
Items to Be Shipped
ITEM
WEIGHT
PROFIT/UNIT
NUMBER AVAILABLE
$3
M2-11
Relationship Between
Items and Stages
ITEM
STAGE
Problem Setup Rollers Air Transport problem is an ideal problem to solve using dynamic programming. Stage 4 will be item 1, stage 3 will be item 2, stage 2 will be item 3,
and stage 1 will be item 4. This is shown in Table M2.3. During the solution, we will be
using stage numbers.
Rollers Air Transport problem can be represented graphically (see Figure M2.8). As
you can see, each item is represented by a stage. Look at stage 4, which is item 1. The
total weight that can be used is represented by s4. This amount is 10 because we havent
assigned any units to be shipped at this time. The decision at this stage is d4 (the number
of units of item 1 we will ship). If d4 is 1, for example, we will be shipping 1 unit of item
1. Also note that r4 is the return or profit at stage 4 (item 1). If we ship 1 unit of item 1, the
profit will be $3.00 (see Table M2.2).
As mentioned previously, the decision variable, dn, represents the number of units of
each item (stage) that can be shipped. Looking back at the original problem, we see that
the problem is constrained by the number of units. This is summarized in the following
table:
STAGE
FIGURE M2.8
Rollers Air Transport Service
Problem
Decisions
d4
s4
Stage 4
(Item 1)
r4
d3
s3
Stage 3
(Item 2)
d2
s2
r3
Stage 2
(Item 3)
r2
Returns
d1
s1
Stage 1
(Item 4)
r1
s0
M2-12
Module 2
DYNAMIC PROGRAMMING
an
bn
cn
1
4
3
2
First note that s4 is 10, the total weight that can be shipped. Because s4 represents the first
item, all 10 tons can be utilized. The transformation equations for the four stages are as
follows:
s3 s4 1d4
stage 4
(a)
s2 s3 4d3
stage 3
(b)
s1 s2 3d2
stage 2
(c)
s0 s1 2d1
stage 1
(d)
Consider stage 3. Equation b reveals that the number of tons still available after stage
3, s2, is equal to the number of tons available before stage 3, s3, minus the number of
tons shipped at stage 3, 4d3. In this equation, 4d3 means that each item at stage 3
weighs 4 tons.
The Return Functions Next, we will look at the return function for each stage. This is
the general form for the return function:
rn (an sn ) (bn dn ) cn
Note that an, bn, and cn are the coefficients for the return function. Using this general form
of the return function, we can put the return function values in the following table:
COEFFICIENTS OF
RETURN FUNCTION
DECISIONS
STAGE
LOWER
UPPER
an
bn
cn
dn
dn
dn
dn
M2-13
The lower value for each decision is zero and the upper value is the total number available. The bn coefficient is the profit per item shipped. The actual return functions are
r4 3d4
r3 9d3
r2 8d2
r1 5d1
Stage-by-Stage Solution As you would expect, the return at any stage, rn, is equal to the
profit per unit at that stage multiplied by the number of units shipped at that stage, dn.
With this information, we can solve Rollers Air Transport problem, starting with stage 1
(item 4). The following table shows the solution for the first stage. You may wish to refer
to Figure M2.8 for this discussion.
STAGE 1
STAGE 1
S1
d1
r1
s0
f0
f1
S1
d1
r1
s0
f0
f1
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
Because we dont know how many tons will be available at stage 1, we must consider
all possibilities. Thus the number of tons available at stage 1, s1, can vary from 1 to 10.
This is seen in the first column of numbers for stage 1. The number of units that we ship at
stage 1, d1, can vary from 0 to 2. We cant go over 2 because the number available is only
2. For any decision we compute the return at stage 1, r1, by multiplying the number of
items shipped by 5, the profit per item. The profit at this stage will be 0, 5, or 10, depending on whether 0, 1, or 2 items are shipped. Note that the total return at this stage, f1, is the
same as r1 because this is the only stage we are considering so far. Also note that the total
M2-14
Module 2
DYNAMIC PROGRAMMING
return before stage 1, f0, is 0 because this is the beginning of the solution and we are shipping nothing at this point.
The solution for stage 1 shows the best decision, the return for this stage, and the total
return given all possible number of tons available (0 to 10 tons). Using the results of stage
1, we can now proceed to stage 2. The solution for stage 2 is as follows:
STAGE 2
STAGE 2
S2
d2
r2
s1
f1
f2
S2
d2
r2
s1
f1
f2
10
10
10
18
16
16
10
10
10
18
10
10
16
21
10
10
10
10
10
18
13
16
21
10
10
10
10
10
13
10
18
16
16
16
10
26
10
The solution for stage 2 is found in exactly the same way as for stage 1. At stage 2 we
still have to consider all possible number of tons available (from 0 to 10). See the s2 column (the first column). At stage 2 (item 3) we still only have 2 units that can be shipped.
Thus d2 (second column) can range from 0 to a maximum of 2. The return for each s2 and
d2 combination at stage 2, r2, is shown in the third column. These numbers are the profit
per item at this stage, 8, times the number of items shipped. Because items shipped at
stage 2 can be 0, 1, or 2, the profit at this stage can be 0, 8, or 16. The return for stage 2
can also be computed from the return function: r2 8d2.
Now look at the fourth column, s1, which lists the number of items available after
stage 2. This is also the number of items available for stage 1. To get this number, we
have to subtract the number of tons we are shipping at stage 2 (which is the tonnage per
unit times the number of units) from the number of tons available before stage 2 (s2).
Look at the row in which s2 is 6 and d2 is 1. We have 6 tons available before stage 2, and
we are shipping 1 item, which weighs 3 tons. Thus we will have 3 tons still available after
this stage. The s1 values can also be determined using the transformation function, which
is s1 s2 3d2.
The last two columns of stage 2 contain the total return. The return after stage 1 and
before stage 2 is f1. These are the same values that appeared under the f1 column for stage
1. The return after stage 2 is f2. It is equal to the return from stage 2 plus the total return
before stage 2.
M2-15
Stage 3 is solved in the same way as stages 1 and 2. The following table presents the
solution for stage 3; look at each row and make sure that you understand the meaning of
each value.
STAGE 3
STAGE 3
s3
d3
r3
s2
f2
f3
10
10
13
13
16
16
14
s3
7
10
d3
r3
s2
f2
f3
18
18
17
21
21
10
19
21
21
13
22
10
26
26
16
25
Now we solve the last stage of the problem, stage 4. The following table shows the
solution procedure:
STAGE 4
S4
d4
r4
s3
f3
f4
10
10
26
26
22
25
21
27
18
27
12
16
28
15
13
28
18
10
28
The first thing to note is that we only have to consider one value for s4, because we
know the number of tons available for stage 4; s4 must be 10 because we have all 10 tons
available. There are six possible decisions at this stage, or six possible values for d4, because the number of available units is 6. The other columns are computed in the same
way. Note that the return after stage 4, f4, is the total return for the problem. We see that
the highest profit is 28. We also see that there are three possible decisions that will give
M2-16
Module 2
DYNAMIC PROGRAMMING
this level of profit, shipping 4, 5, or 6 items. Thus we have alternate optimal solutions.
One possible solution is as follows:
FINAL SOLUTION
STAGE
OPTIMAL DECISION
OPTIMAL RETURN
18
10
Total
28
We start with shipping 6 units at stage 4. Note that s3 is 4 from the stage 4 calculations,
given that d4 is 6. We use this value of 4 and go to the stage 3 calculations. We find the rows
in which s3 is 4 and pick the row with the highest total return, f3. In this row d3 is 0 items
with a total return (f3) of 10. As a result, the number of units available, s2, is still 4. We then
go to the calculations for stage 2 and then stage 1 in the same way. This gives us the optimal
solution already described. See if you can find one of the alternate optimal solutions.
Summary
Dynamic programming is a flexible, powerful technique. A large number of problems can
be solved using this technique, including the shortest-route and knapsack problems. The
shortest-route problem finds the path through a network that minimizes total distance
traveled, while the knapsack problem maximizes value or return. An example of a knapsack problem is to determine what to ship in a cargo plane to maximize total profits given
the weight and size constraints of the cargo plane. The dynamic programming technique
requires four steps: (1) Divide the original problem into stages, (2) solve the last stage
first, (3) work backward solving each subsequent stage, and (4) obtain the optimal solution after all stages have been solved.
Dynamic programming requires that we specify stages, state variables, decision variables, decision criteria, an optimal policy, and a transformation function for each specific
problem. Stages are logical subproblems. State variables are possible input values or beginning conditions. The decision variables are the alternatives that we have at each stage,
such as which route to take in a shortest-route problem. The decision criterion is the objective of the problem, such as finding the shortest route or maximizing total return in a
knapsack problem. The optimal policy helps us obtain an optimal solution at any stage,
and the transformation function allows us to go from one stage to the next.
Glossary
Dynamic Programming. A quantitative technique that works backward from the end of the problem to the beginning of the problem in determining the best decision for a number of interrelated
decisions.
Stage. A logical subproblem in a dynamic programming problem.
State Variable. A term used in dynamic programming to describe the possible beginning situations
or conditions of a stage.
Solved Problem
M2-17
Decision Variable. The alternatives or possible decisions that exist at each stage of a dynamic programming problem.
Decision Criterion. A statement concerning the objective of a dynamic programming problem.
Optimal Policy. A set of decision rules, developed as a result of the decision criteria, that gives optimal decisions at any stage of a dynamic programming problem.
Transformation. An algebraic statement that shows the relationship between stages in a dynamic
programming problem.
Key Equations
(M2-1) sn Input to stage n
The input to stage n. This is also the output from stage n 1.
(M2-2) dn Decision at stage n
The decision at stage n.
(M2-3) rn Return at stage n
The return function, usually profit or loss, at stage n.
(M2-4) sn1 Output from stage n
The output from stage n. This is also the input to stage n 1.
(M2-5) tn Transformation function at stage n
A transformation function that allows us to go from one stage to another.
(M2-6) sn1 tn(sn , dn)
The general relationship that shows how the output from any stage is a function of the
input to the stage and the decisions made at that stage.
(M2-7) fn Total return at stage n
This equation gives the total return (profit or costs) at any stage. It is obtained by
summing the return at each stage, rn.
Solved Problem
Solved Problem M2-1
Lindsey Cortizan would like to travel from the university to her hometown over the holiday season. The road map from the university (node 1) to her home (node 7) is shown in Figure M2.9.
What is the best route that will minimize total distance traveled?
FIGURE M2.9
Road Map for Lindsey Cortizan
39
20
6
13
28
1
18
3
36
10
22
18
2
M2-18
Module 2
DYNAMIC PROGRAMMING
FIGURE M2.10
Solution for the Lindsey
Cortizan Problem
52
4
39
13
20
50
1
41
6
13
28
18
3
36
10
22
18
2
10
28
stage 3
stage 2
stage 1
Solution
The solution to this problem is identical to the one presented earlier in the chapter. First, the
problem is broken into three stages. See the network in Figure M2.10. Working backward, we
start by solving stage 3. The closest and only distance from node 6 to node 7 is 13, and the closest and only distance from node 5 to node 7 is 10. We proceed to stage 2. The minimum distances are 52, 41, and 28 from node 4, node 3, and node 2 to node 7. Finally, we complete stage
3. The optimal solution is 50. The shortest route is 1257 and is shown in the following network. This problem can also be solved using tables, as shown following the network solution.
Problem type
Number of stages
Transition function type
Recursion function type
minimization network
3
sn1 sn dn
fn rn fn1
STAGE
NUMBER OF DECISIONS
STAGE
STARTING NODE
ENDING NODE
RETURN VALUE
22
18
20
18
36
28
39
10
13
Solved Problem
STAGE 1
s1
d1
r1
s0
f0
f1
67
13
13
57
10
10
STAGE 2
s2
d2
r2
s1
f1
f2
46
39
13
52
36
28
13
41
26
36
13
49
25
18
10
28
STAGE 3
s3
d3
r3
s2
f2
f3
14
20
51
71
13
18
41
59
12
22
28
50
FINAL SOLUTION
STAGE
OPTIMAL DECISION
OPTIMAL RETURN
12
22
25
18
57
10
Total
50
M2-19
M2-20
Module 2
DYNAMIC PROGRAMMING
SELF-TEST
Before taking the self-test, refer back to the learning objectives at the beginning of the
module, the notes in the margins, and the glossary at the end of the module.
Use the key at the back of the book to correct your answers.
Restudy pages that correspond to any questions that you answered incorrectly or material
you feel uncertain about.
M2-2
M2-3
M2-4
M2-5
Problems
M2-6 Refer to Figure M2.1. What is the shortest route between Rice and Dixieville if the
road between Hope and Georgetown is improved and the distance is reduced to 4
miles?
M2-7
Due to road construction between Georgetown and Dixieville, a detour must be taken
through country roads (Figure M2.1). Unfortunately, this detour has increased the distance from Georgetown to Dixieville to 14 miles. What should George do? Should he
take a different route?
M2-8
The Rice Brothers have a gold mine between Rice and Brown. In their zeal to find
gold, they have blown up the road between Rice and Brown. The road will not be in
service for five months. What should George do? Refer to Figure M2.1.
M2-9
FIGURE M2.11
(for Problem M2-9)
4
2
10
4
2
7
6
12
5
4
9
6
11
3
4
8
10
6
M2-21
M2-22
Module 2
DYNAMIC PROGRAMMING
10
5
3
3
2
11
4
7
M2-11 Mail Express, an overnight mail service, delivers mail to customers throughout the
United States, Canada, and Mexico. Fortunately, Mail Express has additional capacity
on one of its cargo planes. To maximize profits, Mail Express takes shipments from
local manufacturing plants to warehouses for other companies. Currently, there is
room for another 6 tons. The following table shows the items that can be shipped,
their weights, the expected profit for each, and the number of available parts. How
many units of each item do you suggest that Mail Express ship?
ITEMS TO BE SHIPPED
ITEM
WEIGHT
(TONS)
PROFIT/UNIT
NUMBER
AVAILABLE
$3
M2-12 Leslie Bessler must travel from her hometown to Denver to see her friend Austin. Given
the road map of Figure M2.13, what route will minimize the distance that she travels?
FIGURE M2.13
(for Problem M2-12)
6
2
4
1
11
3
12
4
5
3
10
2
7
M2-13 An air cargo company has the following shipping requirements. Two planes are available with a total capacity of 11 tons. How many of each item should be shipped to
maximize profits?
ITEMS TO BE SHIPPED
ITEM
WEIGHT
(TONS)
PROFIT/UNIT
NUMBER
AVAILABLE
$3
M2-14 Because of a new manufacturing and packaging procedure, the weight of item 2 in
Problem M2-13 can be cut in half. Does this change the number or types of items that
can be shipped by the air transport company?
M2-15 What is the shortest route through the network of Figure M2.14?
3
3
13
16
3
11
15
6
19
2
4
5
4
2
20
3
10
3
3
18
3
7
4
12
4
4
FIGURE M2.14
(for Problem M2-15)
17
14
M2-16 The road between node 6 and node 11 is no longer in service due to construction.
(Refer to Problem M2-15.) What is the shortest route given this situation?
M2-23
M2-24
Module 2
DYNAMIC PROGRAMMING
Case Study
United Trucking
Like many trucking operations, United Trucking got started with
one truck and one ownerJudson Maclay. Judson is an individualist and always liked to do things his way. He was a fast driver,
and many people called the 800 number on the back of his truck
when he worked for Hartmann Trucking. After two years with
Hartmann and numerous calls about his bad driving, Judson decided to go out on his own. United Trucking was the result.
In the early days of United Trucking, Judson was the only
driver. On the back of his truck was the message: How do you
like my driving? Call 1-800-AMI-FAST. He was convinced
that some people actually tried to call the number. Soon, a
number of truck operators had the same or similar messages
on the back of their trucks. After three years of operation, Judson had 15 other trucks and drivers working for him. He
traded his driving skills for office and management skills. Although 1-800-AMI-FAST was no longer visible, Judson decided to never place an 800 number on the back of any of his
trucks. If someone really wanted to complain, they could look
up United Trucking in the phone book.
Judson liked to innovate with his trucking company. He
knew that he could make more money by keeping his trucks
full. Thus he decided to institute a discount trucking service.
He gave substantial discounts to customers that would accept
delivery to the West Coast within two weeks. Customers got a
great price, and he made more money and kept his trucks full.
Over time, Judson developed steady customers that would
usually have loads to go at the discounted price. On one ship-
WEIGHT
(TONS)
PROFIT/UNIT
AVAILABLE
$10
10
20
25
11
30
50
10
10
Discussion Questions
1. What do you recommend Judson should do?
2. If the total available capacity was 20 tons, how would
this change Judsons decision?
Bibliography
Bellman, R. E. Dynamic Programming. Princeton,
NJ: Princeton University Press, 1957.
Bibliography
M2-25
M O D U L E
Decision Theory
and the Normal
Distribution
LEARNING OBJECTIVES
After completing this module, students will be able to:
1. Understand how the normal curve can be used in performing break-even
analysis.
2. Compute the expected value of perfect information (EVPI) using the normal
curve.
3. Perform marginal analysis where products have a constant marginal profit and
loss.
MODULE OUTLINE
M3.1 Introduction
M3.2 Break-Even Analysis and the Normal Distribution
M3.3 EVPI and the Normal Distribution
Summary Glossary Key Equations Solved
Problems Self-Test Discussion Questions and
Problems Bibliography
M3-1
M3-2
Module 3
DECISION THEORY
AND THE
NORMAL DISTRIBUTION
M3.1 INTRODUCTION
In Chapters 3 and 4 in your text we looked at examples that dealt with only a small number of states of nature and decision alternatives. But what if there were 50, 100, or even
1,000s of states and/or alternatives? If you used a decision tree or decision table, solving
the problem would be virtually impossible. This module shows how decision theory can
be extended to handle problems of such magnitude.
We begin with the case of a firm facing two decision alternatives under conditions of
numerous states of nature. The normal probability distribution, which is widely applicable
in business decision making, is first used to describe the states of nature.
Break-even analysis, often called cost-volume analysis, answers several common management questions relating the effect of a decision to overall revenues or costs. At what point
will we break even, or when will revenues equal costs? At a certain sales volume or demand level, what revenues will be generated? If we add a new product line, will this action increase revenues? In this section we look at the basic concepts of break-even
analysis and explore how the normal probability distribution can be used in the decisionmaking process.
$36,000
$36,000
$10 $4
$6
For a detailed explanation of the break-even equation, see Appendix M3.1 at the end of this module.
M3-3
Any demand for the new game that exceeds 6,000 units will result in a profit,
whereas a demand less than 6,000 units will cause a loss. For example, if it turns out that
demand is 11,000 games of Strategy, Barclays profit would be $30,000.
Revenue (11,000 games $10/game)
$110,000
Less expenses
Fixed cost
Variable cost
(11,000 games $4/game)
$36,000
$44,000
Total expense
$ 80,000
Profit
$ 30,000
If demand is exactly 6,000 games (the break-even point), you should be able to compute
for yourself that profit equals $0.
Rudy Barclay now has one useful piece of information that will help him make the
decision about introducing the new product. If demand is less than 6,000 units, a loss will
be incurred. But actual demand is not known. Rudy decides to turn to the use of a probability distribution to estimate demand.
Standard Deviation of
Demand Describes Spread
FIGURE M3.1
Shape of a Typical Normal
Distribution
M3-4
Module 3
DECISION THEORY
AND THE
NORMAL DISTRIBUTION
FIGURE M3.2
Normal Distribution for
Barclays Demand
5,000
8,000
11,000
X
Demand (Games)
Because this is a symmetric distribution, Rudy decides that a normal curve is appropriate. In Chapter 2, we saw how to take the data in a normal curve such as Figure M3.2
and compute the value of the standard deviation. The formula for calculating the number
of standard deviations that any value of demand is away from the mean is
Z
demand m
s
(M3-2)
where Z is the number of standard deviations above or below the mean, m. It is provided
in the table in Appendix A at the end of this text.
We see that the area under the curve to the left of 11,000 units demanded is 85%
of the total area, or 0.85. From Appendix A, the Z value for 0.85 is approximately 1.04.
This means that a demand of 11,000 units is 1.04 standard deviations to the right of the
mean, m.
With m 8,000, Z 1.04, and a demand of 11,000, we can easily compute s.
Z
demand m
s
or
1.04
11,000 8,000
s
or
1.04s 3,000
or
s
3,000
2,885 units
1.04
At last, we can state that Barclays demand appears to be normally distributed, with a
mean of 8,000 games and a standard deviation of 2,885 games. This allows us to answer
some questions of great financial interest to management, such as what the probability is
of breaking even. Recalling that the break-even point is 6,000 games of Strategy, we must
find the number of standard deviations from 6,000 to the mean.
Z
break-even point m
s
6,000 8,000
2,000
0.69
2,885
2,885
M3-5
FIGURE M3.3
Probability of Breaking Even for
Barclays New Game
Loss Area 24.51%
Break-Even
6,000 units
This is represented in Figure M3.3. Because Appendix A is set up to handle only positive
Z values, we can find the Z value for 0.69, which is 0.7549 or 75.49% of the area under
the curve. The area under the curve for 0.69 is just 1 minus the area computed for
0.69, or 1 0.7549. Thus, 24.51% of the area under the curve is to the left of the breakeven point of 6,000 units. Hence,
P(loss) P(demand break-even) 0.2451
24.51%
P(profit) P(demand break-even) 0.7549
75.49%
The fact that there is a 75% chance of making a profit is useful management information
for Rudy to consider.
Before leaving the topic of break-even analysis, we should point out two caveats:
1. We have assumed that demand is normally distributed. If we should find that this is
not reasonable, other distributions may be applied. These are beyond the scope of
this book.
2. We have assumed that demand was the only random variable. If one of the other
variables (price, variable cost, or fixed costs) were a random variable, a similar
procedure could be followed. If two or more variables are both random, the
mathematics becomes very complex. This is also beyond our level of treatment.
(M3-3)
Computing EMV.
M3-6
Module 3
DECISION THEORY
AND THE
NORMAL DISTRIBUTION
Rudy has two choices at this point. He can recommend that the firm proceed with the
new game; if so, he estimates there is a 75% chance of at least breaking even and an EMV
of $12,000. Or, he might prefer to do further marketing research before making a decision. This brings up the subject of the expected value of perfect information.
Lets return to the Barclay Brothers problem to see how to compute the expected value of
perfect information (EVPI) and expected opportunity loss (EOL) associated with introducing the new game. The two steps follow:
X)
$6(6,000
$0
K$0(break-even point X)
where
K loss per unit when sales are below the break-even point
X sales in units
(M3-4)
M3-7
(M3-5)
where
EOL expected opportunity loss
K loss per unit when sales are below the break-even point
s standard deviation of the distribution
D
m break-even point
s
(M36)
where
absolute value sign
m mean sales
N(D) value for the unit normal loss integral in Appendix B for a given value of D
Here is how Rudy can compute EOL for his situation:
K $6
s 2,885
D
8,000 6,000
0.69 0.60 0.09
2,885
Now refer to the unit normal loss integral table. Look in the 0.6 row and read over to
the 0.09 column. This is N(0.69), which is 0.1453.
N(0.69) 0.1453
Therefore,
EOL KN(0.69)
($6)(2,885)(0.1453) $2,515.14
Because EVPI and EOL are equivalent, the expected value of perfect information is
also $2,515.14. This is the maximum amount that Rudy should be willing to spend on additional marketing information.
The relationship between the opportunity loss function and the normal distribution is
shown in Figure M3.4. This graph shows both the opportunity loss and the normal distribution with a mean of 8,000 games and a standard deviation of 2,885. To the right of the
break-even point we note that the loss function is 0. To the left of the break-even point,
M3-8
Module 3
DECISION THEORY
FIGURE M3.4
Barclays Opportunity Loss
Function
AND THE
NORMAL DISTRIBUTION
8,000 Games
Loss ($)
Normal Distribution
8,000
2,885
Slope 6
6,000
X
Demand (Games)
the opportunity loss function increases at a rate of $6 per unit, hence the slope of 6. The
use of Appendix B and Equation M3-5 allows us to multiply the $6 unit loss times each of
the probabilities between 6,000 units and 0 units and to sum these multiplications.
Summary
In this module we looked at decision theory problems that involved many states of nature
and alternatives. As an alternative to decision tables and decision trees, we learned to use
the normal distribution to solve break-even problems and find the expected monetary
value and EVPI. We need to know the mean and standard deviation of the normal distribution and to be certain that it is the appropriate probability distribution to apply. Other
continuous distributions can also be used, but they are beyond the level of this book.
Glossary
Break-Even Analysis. The analysis of relationships between profit, costs, and demand level.
Opportunity Loss Function. A function that relates opportunity loss in dollars to sales in units.
Unit Normal Loss Integral. A table that is used in the determination of EOL and EVPI.
Key Equations
(M3-1) Break-even point (in units)
fixed cost
price/unit variable cost/unit
The formula that provides the volume at which total revenue equals total costs.
(M3-2) Z
demand m
s
Solved Problems
K (break-even point X)
for X break-even point
$0
for X break-even point
m break-even point
s
Solved Problems
Solved Problem M3-1
Terry Wagner is considering self-publishing a book on yoga. She has been teaching yoga for
more than 20 years. She believes that the fixed costs of publishing the book will be about
$10,000. The variable costs are $5.50, and the price of the yoga book to bookstores is expected
to be $12.50. What is the break-even point for Terry?
Solution
This problem can be solved using the break-even formulas in the supplement, as follows:
Break-even point (BEP) in units
BEP
$10,000
$12.50 $5.50
$10,000
$7
1,429 units
Solved Problem M3-2
In this supplement we discussed how we could use the normal curve to determine the expected
opportunity loss (EOL). We have determined that D 0.70, the standard deviation is 1,500, and
K is $10. Given these data, determine EOL.
Solution
The first step is to go to Appendix B and get a value of N(D) when D is 0.7. Looking at Appendix B for the unit normal loss integral, we see that N(D) is 0.1429. This value can be placed into
the equation for EOL as follows:
EOL $10(1,500)(0.1429) $2,143.50
M3-9
M3-10
Module 3
DECISION THEORY
AND THE
NORMAL DISTRIBUTION
SELF-TEST
Before taking the self-test, refer back to the learning objectives at the beginning of the
module and the glossary at the end of the module.
Use the key at the back of the book to correct your answers.
Restudy pages that correspond to any questions that you answered incorrectly or material
you feel uncertain about.
Problems
M3-5
M3-11
M3-12
Module 3
DECISION THEORY
AND THE
NORMAL DISTRIBUTION
M3-11 Elizabeth Belt is not completely certain that the loss per lamp is $10 if sales are below
the break-even point (refer to Problem M3-10). The loss per lamp could be as low as
$8 or as high as $15. What effect would these two values have on the expected opportunity loss?
M3-12 Leisure Supplies is considering the possibility of using a new process for producing
sinks. This new process would increase the fixed cost by $16,000. In other words, the
fixed cost would double (see Problem M3-9). This new process will improve the quality of the sinks and reduce the cost it takes to produce each sink. It will cost only $19
to produce the sinks using the new process.
(a) What do you recommend?
(b) Leisure Supplies is considering the possibility of increasing the purchase price to
$32 using the old process given in Problem M3-9. It is expected that this will lower
the mean sales to 26,000 units. Should Leisure Supplies increase the selling price?
M3-13 Quality Cleaners specializes in cleaning apartment units and office buildings. Although the work is not too enjoyable, Joe Boyett has been able to realize a considerable profit in the Chicago area. Joe is now thinking about opening another Quality
Cleaners in Milwaukee. To break even, Joe would need to get 200 cleaning jobs per
year. For every job under 200, Joe will lose $80. Joe estimates that the average sales
in Milwaukee are 350 jobs per year, with a standard deviation of 150 jobs. A marketing research team has approached Joe with a proposition to perform a marketing
study on the potential for his cleaning business in Milwaukee. What is the most that
Joe would be willing to pay for the marketing research?
M3-14 Diane Kennedy is contemplating the possibility of going into competition with Primary Pumps, a manufacturer of industrial water pumps. Diane has gathered some
interesting information from a friend of hers who works for Primary. Diane has
been told that the mean sales for Primary are 5,000 units and the standard deviation is 50 units. The opportunity loss per pump is $100. Furthermore, Diane has
been told that the most that Primary is willing to spend for marketing research for
the demand potential for pumps is $500. Diane is interested in knowing the breakeven point for Primary Pumps. Given this information, compute the break-even
point.
M3-15 Jack Fuller estimates that the break-even point for EM 5, a standard electrical motor, is
500 motors. For any motor that is not sold, there is an opportunity loss of $15. The average sales have been 700 motors, and 20% of the time sales have been between 650 and
750 motors. Jack has just been approached by Radner Research, a firm that specializes in
performing marketing studies for industrial products, to perform a standard marketing
study. What is the most that Jack would be willing to pay for the marketing research?
M3-16 Jack Fuller believes that he has made a mistake in his sales figures for EM 5 (see
Problem M3-15 for details). He believes that the average sales are 750 instead of 700
units. Furthermore, he estimates that 20% of the time, sales will be between 700 and
800 units. What effect will these changes have on your estimate of the amount that
Jack should be willing to pay for the marketing research?
Bibliography
Drenzer and Wesolowsky. The Expected Value of
Perfect Information in Facility Location,
Operations Research (MarchApril 1980): 395.
Hammond, J. S., R. L. Kenney, and H. Raiffa. The
Hidden Traps in Decision Making, Harvard
fixed cost
price/unit variable cost/unit
M3-13
M O D U L E
Material Requirements
Planning and Just-inTime Inventory
LEARNING OBJECTIVES
After completing this module, you will be able to:
1. Describe the use of material requirements planning (MRP) in solving
dependent-demand inventory problems.
MODULE OUTLINE
M4.1 Introduction
M4.2 Dependent Demand: The Case for Material Requirements
Planning
M4-1
M4-2
Module 4
AND
J U S T - I N -T I M E I N V E N T O R Y
M4.1 INTRODUCTION
REQUIREMENTS PLANNING
In all the inventory models we discussed in Chapter 6, we assumed that the demand for
one item was independent of the demand for other items. For example, the demand for refrigerators is usually independent of the demand for toaster ovens. Many inventory problems, however, are interrelated; the demand for one item is dependent on the demand for
another item. Consider a manufacturer of small power lawn mowers. The demand for
lawn mower wheels and spark plugs is dependent on the demand for lawn mowers. Four
wheels and one spark plug are needed for each finished lawn mower. Usually when the
demand for different items is dependent, the relationship between the items is known and
constant. Thus, you should forecast the demand for the final products and compute the requirements for component parts.
As with the inventory models discussed previously, the major questions that must be
answered are how much to order and when to order. But with dependent demand, inventory scheduling and planning can be very complex indeed. In these situations, MRP can be
employed effectively. Some of the benefits of MRP are
1.
2.
3.
4.
5.
6.
Although most MRP systems are computerized, the analysis is straightforward and
similar from one computerized system to the next. Here is the typical procedure.
Step 1 is to develop a material structure tree. Lets say that demand for product A is 50
units. Each unit of A requires 2 units of B and 3 units of C. Now, each unit of B requires 2
units of D and 3 units of E. Furthermore, each unit of C requires 1 unit of E and 2 units of
F. Thus, the demand for B, C, D, E, and F is completely dependent on the demand for A.
Given this information, a material structure tree can be developed for the related inventory
items (see Figure M4.1).
The structure tree has three levels: 0, 1, and 2. Items above any level are called parents, and items below any level are called components. There are three parents: A, B, and
C. Each parent item has at least one level below it. Items B, C, D, E, and F are compo-
FIGURE M4.1
Material Structure Tree for
Item A
Level
0
M4-3
B(2)
D(2)
C(3)
E(3)
E(1)
F(2)
nents because each item has at least one level above it. In this structure tree, B and C are
both parents and components.
Note that the number in the parentheses in Figure M4.1 indicates how many units of
that particular item are needed to make the item immediately above it. Thus B(2) means
that it takes 2 units of B for every unit of A, and F(2) means that it takes 2 units of F for
every unit of C.
After the material structure tree has been developed, the number of units of each item
required to satisfy demand can be determined. This information can be displayed as follows:
Part B:
2 number of As 2 50 100
Part C:
3 number of As 3 50 150
Part D:
Part E:
3 number of Bs
1 number of Cs 3 100 1 150 450
Part F:
Thus, for 50 units of A we need 100 units of B, 150 units of C, 200 units of D, 450
units of E, and 300 units of F. Of course, the numbers in this table could have been determined directly from the material structure tree by multiplying the numbers along the
branches times the demand for A, which is 50 units for this problem. For example, the
number of units of D needed is simply 2 2 50 200 units.
M4-4
Module 4
FIGURE M4.2
Gross Material Requirements
Plan for 50 Units of A
AND
J U S T - I N -T I M E I N V E N T O R Y
Week
1
50
Required Date
A
Order Release
50
Required Date
100
100
Order Release
150
Required Date
C
150
Order Release
200
Required Date
200
Order Release
300
Required Date
150
E
300
Order Release
150
300
Required Date
F
300
Order Release
one week to make C, one week to make D, two weeks to make E, and three weeks to
make F. With this information, the gross material requirements plan can be constructed to
reveal the production schedule needed to satisfy the demand of 50 units of A at a future
date. Refer to Figure M4.2.
The interpretation of the material in Figure M4.2 is as follows: If you want 50 units
of A at week 6, you must start the manufacturing process in week 5. Thus, in week 5 you
need 100 units of B and 150 units of C. These two items take 2 weeks and 1 week to produce. (See the lead times.) Production of B should be started in week 3, and C should be
started in week 4. (See the order release for these items.) Working backward, the same
computations can be made for all the other items. The material requirements plan graphically reveals when each item should be started and completed in order to have 50 units of
A at week 6. Now, a net requirements plan can be developed given the on-hand inventory
in Table M4.1; here is how it is done.
Using these data, we can develop a net material requirements plan that includes gross
requirements, on-hand inventory, net requirements, planned-order receipts, and plannedorder releases for each item. It is developed by beginning with A and working backward
through the other items. Figure M4.3 shows a net material requirements plan for product A.
The net requirements plan is constructed like the gross requirements plan. Starting
with item A, we work backward determining net requirements for all items. These computations are done by referring constantly to the structure tree and lead times. The gross
requirements for A are 50 units in week 6. Ten items are on hand, and thus the net requirements and planned-order receipt are both 40 items in week 6. Because of the oneweek lead time, the planned-order release is 40 items in week 5. (See the arrow
TABLE M4.1
On-Hand Inventory
ITEM
ON-HAND INVENTORY
10
15
20
10
10
Week
Item
A
50
10
40
40
Gross
On-Hand 10
Net
Order Receipt
Order Release
80
15
65
65
65
120
20
100
100
Gross
On-Hand 20
Net
Order Receipt
Order Release
100
Gross
On-Hand 10
Net
Order Receipt
Order Release
Gross
On-Hand 5
Net
Order Receipt
Order Release
40
Gross
On-Hand 15
Net
Order Receipt
Order Release
Gross
On-Hand 10
Net
Order Receipt
Order Release
Lead
Time
130
10
120
120
120
195
10
185
185
185
100
0
100
100
100
200
5
195
195
195
M4-5
FIGURE M4.3
Net Material Requirements
Plan for 50 units of A.
M4-6
Module 4
AND
J U S T - I N -T I M E I N V E N T O R Y
connecting the order receipt and order release.) Look down column 5 and refer to the
structure tree in Figure M4.1. Eighty (2 40) items of B and 120 3 40 items of C
are required in week 5 in order to have a total of 50 items of A in week 6. The letter A in
the upper-right corner for items B and C means that this demand for B and C was generated as a result of the demand for the parent, A. Now the same type of analysis is done for
B and C to determine the net requirements for D, E, and F.
3 number of AAs 3 10 30
Part F:
2 number of AAs 2 10 20
AA
D(3)
F(2)
To develop a net material requirements plan, we need to know the lead time for AA. Lets
assume that it is one week. We also assume that we need 10 units of AA in week 6 and
that we have no units of AA on hand.
IN ACTION
Cal Monteith, Compaqs manager of master planning and production control in Houston, was in the process of phasing out
one of Compaqs personal computer models when he was told
that Compaq had underestimated demand. The new schedule
suggested that he build 10,000 more PCs. Could he do it?
Questions Monteith faced were: What parts were on hand and
on order? What labor was available? Could the plant handle
the capacity? Did vendors have the capacity? What product
lines could be rescheduled?
Traditionally, amassing such information required not
only MRP reports but a variety of additional reports. Even
then a response was based on partial information.
Week
Item
Inventory
AA
Gross
On-Hand: 0
Net
Order Receipt
Order Release
1 Week
50
10
40
40
1 Week
40
Gross
On-Hand: 15
Net
Order Receipt
Order Release
80
15
65
65
2 Weeks
65
120
20
100
100
Gross
On-Hand: 20
Net
Order Receipt
Order Release
1 Week
100
130
10
120
120
Gross
On-Hand: 10
Net
Order Receipt
Order Release
Gross
On-Hand: 5
Net
Order Receipt
Order Release
10
0
10
10
10
Gross
On-Hand: 10
Net
Order Receipt
Order Release
Gross
On-Hand: 10
Net
Order Receipt
Order Release
Lead
Time
120
B
100
0
100
100
1 Week
2 Weeks
100
200
5
195
195
195
AA
30
195
10
185
185
185
30
0
30
30
20
0
20
20
AA
3 Weeks
20
Now, we are in a position to modify the net material requirements plan for product A
to include AA. This is done in Figure M4.4.
Look at the top row of the figure. As you can see, we have a gross requirement of 10
units of AA in week 6. We dont have any units of AA on hand, so the net requirement is
also 10 units of AA. Because it takes one week to make AA, the order release of 10 units
of AA is in week 5. This means that we start making AA in week 5 and have the finished
units in week 6.
Because we start making AA in week 5, we must have 30 units of D and 20 units of F
in week 5. See the rows for D and F in Figure M4.4. The lead time for D is one week.
Thus, we must give the order release in week 4 to have the finished units of D in week 5.
Note that there was no inventory on hand for D in week 5. The original 10 units of inventory of D were used in week 5 to make B, which was subsequently used to make A. We
M4-7
FIGURE M4.4
Net Material Requirements
Plan, Including AA
M4-8
Module 4
AND
J U S T - I N -T I M E I N V E N T O R Y
During the past two decades, there has been a trend to make the manufacturing process
more efficient. One objective is to have less in-process inventory on hand. This is known
as just-in-time (JIT) inventory. With this approach, inventory arrives just in time to be
used during the manufacturing process to produce subparts, assemblies, or finished goods.
One technique of implementing JIT is a manual procedure called Kanban. Kanban in
Japanese means card. With a dual-card Kanban system, there is a conveyance Kanban,
or C-Kanban, and a production Kanban, or P-Kanban. The Kanban system is very simple.
Here is how it works:
As seen in Figure M4.5, full containers along with their C-Kanban go from the storage area to a user area, typically on a manufacturing line. During the production process,
parts in the container are used up. When the container is empty, the empty container along
with the same C-Kanban goes back to the storage area. Here, the user picks up a new full
container. The P-Kanban from the full container is removed and sent back to the production area along with the empty container to be refilled.
P = Kanban
and
Container
FIGURE M4.5
The Kanban System
C = Kanban
and
Container
M4-9
Producer
Area
Storage
Area
3
User
Area
2
At a minimum, two containers are required using the Kanban system. One container
is used at the user area, while another container is being refilled for future use. In reality,
there are usually more than two containers. This is how inventory control is accomplished. Inventory managers can introduce additional containers and their associated PKanbans into the system. In a similar fashion, the inventory manager can remove
containers and the P-Kanbans to have tighter control over inventory buildups.
In addition to being a simple, easy-to-implement system, the Kanban system can also
be very effective in controlling inventory costs and in uncovering production bottlenecks.
Inventory arrives at the user area or on the manufacturing line just when it is needed. Inventory does not build up unnecessarily, cluttering the production line or adding to unnecessary
inventory expense. The Kanban system reduces inventory levels and makes for a more effective operation. It is like putting the production line on an inventory diet. Like any diet,
the inventory diet imposed by the Kanban system makes the production operation more
streamlined. Furthermore, production bottlenecks and problems can be uncovered. Many
production managers remove containers and their associated P-Kanban from the Kanban
system in order to starve the production line to uncover bottlenecks and potential problems.
In implementing a Kanban system, a number of work rules or Kanban rules are normally implemented. One typical Kanban rule is that no containers are filled without the
appropriate P-Kanban. Another rule is that each container must hold exactly the specified
number of parts or inventory items. These and similar rules make the production process
more efficient. Only those parts that are actually needed are produced. The production department does not produce inventory just to keep busy. It produces inventory or parts only
when they are needed in the user area or on an actual manufacturing line.
IN ACTION
Source: Holder Roy, Works Management 48, 3 (March 1995): 1821; Works
Management 48, 3 (March 1995): 7; and Jeffrey L. Funk, International Journal
of Operations & Production Management 15, 5 (1995): 6071.
M4-10
Module 4
AND
J U S T - I N -T I M E I N V E N T O R Y
Summary
The demand for inventory is not always independent. When demand is dependent, a technique such as MRP is needed. MRP can be used to determine the gross and net material
requirements for products. This module also investigated the use of JIT inventory. JIT can
lower inventory levels, reduce costs, and make a manufacturing process more efficient.
Kanban, a Japanese word meaning card, is one way to implement the JIT approach.
Glossary
Material Requirements Planning. An inventory model that can handle dependent demand.
Just-in-Time (JIT) Inventory. An approach whereby inventory arrives just in time to be used in
the manufacturing process.
Kanban. A manual JIT system developed by the Japanese. Kanban means card in Japanese.
Solved Problems
Solved Problem M4-1
Because of a technological breakthrough in the product illustrated in Figure M4.1, it now takes
only 2 units of item C to make 1 unit of item A. How does this technological breakthrough
change the material structure tree and the quantities and types of materials required to make 50
units of item A? Assume that there is no on-hand inventory.
Solution
The technological breakthrough changes the structure tree we saw in Figure M4.1 and changes
what materials and quantities are required to make 50 units of item A. These changes are shown
in the following figure.
Material Structure Tree for Item A
A
level
0
B (2)
D(2)
C (2)
E(3)
Part B:
2 number of As 2 50 100
Part C:
2 number of As 2 50 100
Part D:
E(1)
F(2)
Solved Problems
Part E:
3 number of Bs 1 number of Cs
3 100 1 100 400
Part F:
A Gross
LEAD
TIME
50
10
Net
40
Order Receipt
40
a
On-Hand 10
Order Release
40
80A
B Gross
15
Net
65
Order Receipt
65
Order Release
On-Hand 15
65
120A
C Gross
On-Hand 10
10
110
Order Receipt
110
Order Release
110
B
D Gross
130
On-Hand 10
5
125
Order Receipt
125
125
Net
Net
Order Release
M4-11
AND
J U S T - I N -T I M E I N V E N T O R Y
WEEK
ITEM
195B
110C
10
Net
185
110
Order Receipt
185
110
E Gross
Order Release
185
110
220C
F Gross
On-Hand 5
Net
215
Order Receipt
215
Order Release
215
LEAD
TIME
2
On-Hand 10
Module 4
M4-12
Self-Test
M4-13
SELF-TEST
Before taking the self-test, refer back to the learning objectives at the beginning of the
module and the glossary at the end of the module.
Use the key at the back of the book to correct your answers.
Restudy pages that correspond to any questions that you answered incorrectly or material
you feel uncertain about.
M4-14
Module 4
AND
J U S T - I N -T I M E I N V E N T O R Y
M4-2
M4-3
What is the difference between the gross and net material requirements plan?
M4-4
M4-5
Problems
M4-6 This supplement presented a material structure tree for item A in Figure M4-1. Assume
that it now takes 1 unit of item B to make every unit of item A. What impact does this
have on the material structure tree and the number of items of D and E that are needed?
M4-7
Given the information in the Problem M4-6, develop a gross material requirements
plan for 50 units of item A.
M4-8
Using the data from Figures M4-1 through M4-3, develop a net material requirements
plan for 50 units of item A assuming that it only takes 1 unit of item B for each unit of
item A.
M4-9
The demand for product S is 100 units. Each unit of S requires 1 unit of T and Z\x unit
of U. Each unit of T requires 1 unit of V, 2 units of W, and 1 unit of X. Finally, each
unit of U requires Z\x unit of Y and 3 units of Z. All items are manufactured by the
same firm. It takes two weeks to make S, one week to make T, two weeks to make U,
two weeks to make V, three weeks to make W, one week to make X, two weeks to
make Y, and one week to make Z.
(a) Construct a material structure tree and a gross material requirements plan for the
dependent inventory items.
(b) Identify all levels, parents, and components.
(c) Construct a net material requirements plan using the following on-hand inventory
data:
ITEM
ON-HAND INVENTORY
20
20
10
30
30
25
15
10
Bibliography
Bibliography
Allnoch, Allen. Manufacturing Software Plays Key
Role, IE Solutions (November 1997): 1085.
Brucker, H. D., G. A. Flowers, and R. D. Peck.
MRP Shop-Floor Control in a Job Shop:
Definitely Works, Production and Inventory
Management Journal 33, 2 (Second Quarter
1992): 43.
Ding, F., and M. Yuen. A Modified MRP for a
Production System with the Coexistence of
MRP and Kanbans, Journal of Operations
Management 10, 2 (April 1991): 267277.
Fully Automated System Achieves True JIT,
Modern Materials Handling (April 1998): 122.
Funk, Jeffrey L. Just-in-Time Manufacturing and
Logistical Complexity: A Contingency Model,
M4-15
M O D U L E
Mathematical Tools:
Determinants and
Matrices
LEARNING OBJECTIVES
After completing this module, students will be able to:
1. Understand how matrices and determinants are used as mathematical tools in
QA.
2.
3.
4.
5.
MODULE OUTLINE
M5.1 Introduction
M5.2 Determinants
M5.3 Matrices
Summary Glossary Self-Test Problems
Bibliography
M5-1
AND
MATRICES
M5.1 INTRODUCTION
Two new mathematical concepts, determinants and matrices, are introduced in this module. These tools are especially useful in Chapter 16 and the Supplement to Chapter 1,
which deal with Markov analysis and game theory, but they are also handy computational
aids for many other quantitative analysis problems, including linear programming, the
topic of Chapters 7, 8, and 9.
M5.2 DETERMINANTS
A determinant is simply a square array of numbers arranged in rows and columns. Every
determinant has a unique numerical value for which we can solve. As a mathematical tool,
determinants are of value in helping to solve a series of simultaneous equations.
A 2-row by 2-column (2 2) determinant will have the following form, where a, b,
c, and d are numbers:
a b
c d
a
d
g
b c
e f
h i
One common procedure for finding the numerical value of a 2 2 or 3 3 determinant is to draw its primary and secondary diagonals. In the case of a 2 2 determinant,
the value is found by multiplying the numbers on the primary diagonal and subtracting
from that product the product of the numbers on the secondary diagonal:
value (a)(d) (c)(b)
Secondary
diagonal
a b
c d
Primary
diagonal
For a 3 3 determinant, we redraw the first two columns to help visualize all diagonals
and follow a similar procedure.
b
e
h
b c a
e f d
h i g
a
Value
a
d
g
Primary
diagonal
Module 5
a a aa a a
M5-2
Secondary
diagonal
M5.2 Determinants
Lets use this approach to find the numerical values of the following 2 2 and 3 3
determinants:
(a)
3
(b) 2
4
2 5
1 8
2 5
1 8
a
2
1
1
2
1
1
3
2
4
1
5
2
a
1
5
2
3
(b) 2
4
1
5
2
(a)
Value (3) (5) (1) (1) (1) (4) (2) (2) (2)
(4) (5) (2) (2) (1) (3) (1) (2) (1)
15 4 8 40 6 2 51
A set of simultaneous equations may be solved through the use of determinants by
setting up a ratio of two special determinants for each unknown variable. This fairly easy
procedure is best illustrated with an example.
Given the three simultaneous equations,
2X 3Y 1Z 10
4X 1Y 2Z 8
5X 2Y 3Z 6
we may structure determinants to help solve for unknown quantities X, Y, and Z.
3
1
2
1
2
3
Y
2
4
5
2
4
5
10
8
6
1
2
3
2
4
5
3
1
2
1
2
3
1
2
3
3
1
2
X
10
8
6
M5-3
Z
2
4
5
3
1
2
10
8
6
2
4
5
3
1
2
1
2
3
AND
MATRICES
Module 5
M5-4
Determining the values of X, Y, and Z now involves finding the numerical values of
the four separate determinants using the method shown earlier in this supplement.
X
Y
20
0.61
33
Z
134
4.06
33
To verify that X 3.88, Y 0.61, and Z 4.06, we may choose any one of the original
three simultaneous equations and insert these numbers. For example,
2X 3Y 1Z 10
2(3.88) 3(0.61) 1(4.06) 7.76 1.83 4.06
10
M5.3 MATRICES
A matrix, like a determinant, can also be defined as an array of numbers arranged in rows
and columns. Matrices, which are usually enclosed in parentheses or brackets, have no
numerical value as do determinants but are used as an effective means of presenting or
summarizing business data.
The following 2-row by 3-column (2 3) matrix, for example, might be used by
television station executives to describe the channel switching behavior of their 5 oclock
TV news audience.
AUDIENCE SWITCHING PROBABILITIES,
NEXT MONTHS ACTIVITY
CHANNEL
6
CHANNEL
8
STOP
VIEWING
Channel 6
0.80
0.15
0.05
Channel 8
0.20
0.70
0.10
CURRENT
STATION
2 3 matrix
The number in the first row and first column indicates that there is a 0.80 probability
that someone currently watching the Channel 6 news will continue to do so next month.
Similarly, 15% of Channel 6s viewers are expected to switch to Channel 8 next month
M5.3 Matrices
M5-5
(row 1, column 2), 5% will not be watching the 5 oclock news at all (row 1, column 3),
and so on for the second row.
The remainder of this module deals with the numerous mathematical operations that
can be performed on matrices. These include matrix addition, subtraction and multiplication, transposing a matrix, finding its cofactors and adjoint, and matrix inversion.
52 71
3 6
matrix B
3 8
matrix A
1
7
Matrix Multiplication
Matrix multiplication is an operation that may take place only if the number of columns in
the first matrix equals the number of rows in the second matrix. Thus, matrices of the dimensions in the following table may be multiplied:
MATRIX B SIZE
SIZE OF
A B RESULTING
33
33
33
31
13
33
31
11
31
24
43
23
69
92
62
83
36
86
MATRIX A SIZE
M5-6
Module 5
MATRICES
AND
We also note, in the far right column in the table, that the outer two numbers in the matrix
sizes determine the dimensions of the new matrix. That is, if an 8-row by 3-column matrix
is multiplied by a 3-row by 6-column matrix, the resultant product will be an 8-row by 6column matrix.
Matrices of the dimensions in the following table may not be multiplied:
Multiplying numbers.
MATRIX A SIZE
MATRIX B SIZE
34
33
12
12
69
89
22
33
To perform the multiplication process, we take each row of the first matrix and multiply its elements times the numbers in each column of the second matrix. Hence the number in the first row and first column of the new matrix is derived from the product of the
first row of the first matrix times the first column of the second matrix. Similarly, the
number in the first row and second column of the new matrix is the product of the first
row of the first matrix times the second column of the second matrix. This concept is not
nearly as confusing as it may sound.
Let us begin by computing the value of matrix C, which is the product of matrix A
times matrix B:
matrix A
5
2
3
matrix B (4
6)
This is a legitimate task since matrix A is 3 1 and matrix B is 1 2. The product, matrix C, will have 3 rows and 2 columns (3 2).
Symbolically, the operation is matrix A matrix B matrix C:
a
b
c
(d
e)
ad
bd
cd
ae
be
ce
(M5-1)
5
2
3
(4 6)
20
8
12
30
12
18
matrix C
3
1
2
M5.3 Matrices
(3 1)
(1 1)
(a b c)
(ad be cf)
(6 2 5)
d
e
f
3
1
2
67 21
matrix V
matrix U matrix V
(2 2) (2 2)
35 48
matrix Y
(2 2)
af bh
cf dh
24 16
28 8
(M5-2)
2826 4036
To introduce a special type of matrix, called the identity matrix, lets try a final multiplication example:
matrix H
42 73
matrix I
10 01
07
03
42 73
Matrix I is called an identity matrix. An identity matrix has 1s on its diagonal and 0s
in all other positions. When multiplied by any matrix of the same square dimensions, it
yields the original matrix. So in this case, matrix J matrix H.
Matrix multiplication can also be useful in performing business computations.
Blank Plumbing and Heating is about to bid on three contract jobs: to install plumbing fixtures in a new university dormitory, an office building, and an apartment complex.
M5-7
M5-8
Module 5
AND
MATRICES
The number of toilets, sinks, and bathtubs needed at each project is summarized in matrix
notation as follows. The cost per plumbing fixture is also given. Matrix multiplication
may be used to provide an estimate of total cost of fixtures at each job.
PROJECT
DEMAND
COST/UNIT
Toilets
Sinks
Bathtubs
Dormitory
10
Toilet
$40
Office
20
20
Sink
$25
Apartments
15
30
15
Bathtub
$50
5
20
15
(3 3)
(3 1)
10
20
30
2
0
15
(3 1)
$40
$25
$50
$550
$1,300
$2,100
Hence Blank Plumbing can expect to spend $550 on fixtures at the dormitory project,
$1,300 at the office building, and $2,100 at the apartment complex.
Matrix Transpose
The transpose of a matrix is a means of presenting data in different form. To create the
transpose of a given matrix, we simply interchange the rows with the columns. Hence, the
first row of a matrix becomes its first column, the second row becomes the second column, and so on.
Two matrices are transposed here:
matrix A
transpose of matrix A
matrix B
transpose of matrix B
2
0
4
6
9
8
5
2
6
3
0
9
1
4
8
28
7
5
0
6
5
3
1
2
7
0
3
8
5
6
4
3
4
M5.3 Matrices
Lets compute the matrix of cofactors, and then the adjoint, for the following matrix,
using Table M5.1 on the next page to help in the calculations:
original matrix
7
0
1
matrix of cofactors
3
51
21
4
4
1
2
25
14
3
4
2
51
4
25
21
1
14
3
2
4
5
3
8
(M5-3)
M5-9
Module 5
AND
MATRICES
ELEMENT
REMOVED
Row 1, column 1
01 38
Row 1, column 2
24 38
Row 1, column 3
24 01
Row 2, column 1
71 58
Row 2, column 2
34 58
Row 2, column 3
34 71
Row 3, column 1
70 53
Row 3, column 2
32 53
32 70
VALUE OF
COFACTOR
0
1
3
3
8
2
4
3
8
4 (sign changed)
2
4
0
1
7
1
5
8
51
3
4
5
8
3
4
7
25
1
25 (sign changed)
7
0
5
3
3
2
5
1
3
1 (sign changed)
3
2
7 14
0
21
The adjoint of a matrix is extremely helpful in forming the inverse of the original matrix. We simply compute the value of the determinant of the original matrix and divide
each term of the adjoint by this value.
To find the inverse of the matrix just presented, we need to know the adjoint (already
computed) and the value of the determinant of the original matrix:
3
2
4
7
0
1
5
3
8
original matrix
Value of determinant:
3 7
2 0
4 1
a
5
3
8
7
0
1
3
2
4
Row 3, column 3
DETERMINANT
OF COFACTORS
COFACTORS
TABLE M5.1
M5-10
Glossary
51
21
4
4
27
2
25
27
14
27
inverse
27
27
27
3
51
4
4
27
1
2
25
27
14
27
27
27
27
27
21
27
27
27
27
27
We may verify that this is indeed the correct inverse of the original matrix by multiplying
the original matrix times the inverse:
original
matrix
3
2
4
7
0
1
5
3
8
3
27 5127
4
27 427
2
27 2527
identity
matrix
inverse
21
27
27
14
27
1
1
0
0
0
1
0
0
0
1
Summary
This module contained a brief presentation of determinants and matrices, two mathematical tools often used in quantitative analysis. Determinants are useful in solving a series of
simultaneous equations. Matrices are the basis for the simplex method of linear programming. The modules discussion included matrix addition, subtraction, multiplication,
transposition, cofactors, adjoints, and inverses.
Glossary
Determinant. A square array of numbers arranged in rows and columns. Every determinant has a
unique numerical value.
Simultaneous Equations. A series of equations that must be solved at the same time.
Matrix. An array of numbers that can be used to present or summarize business data.
Identity Matrix. A square matrix with 1s on its diagonal and 0s in all other positions.
Transpose. The interchange of rows and columns in a matrix.
Matrix of Cofactors. The determinants of the numbers remaining in a matrix after a given row and
column have been removed.
Adjoint. The transpose of a matrix of cofactors.
Inverse. A unique matrix that may be multiplied by the original matrix to create an identity matrix.
M5-11
M5-12
Module 5
AND
MATRICES
SELF-TEST
Before taking the self-test, refer back to the learning objectives at the beginning of the
module and the glossary at the end of the module.
Use the key at the back of the book to correct your answers.
Restudy pages that correspond to any questions that you answered incorrectly or material
you feel uncertain about.
1. A determinant is ________________ .
2. The value of the determinant
1
3
2
is
4
a. 2.
b. 10.
c. 2.
d. 5.
e. 14.
3. To find the value of a small determinant, you
________________ .
4. Matrices
a. are usually enclosed in parentheses or brackets.
b. can be defined as an array of numbers in rows and
columns.
c. have no numerical value.
d. are an effective way of presenting or summarizing
data.
e. all of the above.
Problems
Problems
M5-1 Find the numerical values of the following determinants.
(a)
6
5
3
2
3
(b) 1
4
7
1
3
6
2
2
matrix A
matrix B
matrix D
5
2
6
1
1
7
matrix C
4
8
2
3
7
0
3
7
9
6
8
2
9
1
4
1
0
1
6
6
5
5)
5
4
3
matrix A
21
matrix E (5
matrix B (3
1) matrix F
4
3
2
0
matrix R
21 34
matrix S
10 01
matrix W
matrix Y
12
3
2
4
5
1
4
4
3
5
6
1
5
M5-13
M5-14
Module 5
M5-5
AND
MATRICES
RLB Electrical Contracting, Inc., bids on the same three jobs as Blank Plumbing
(Section M5.3). RLB must supply wiring, conduits, electrical wall fixtures, and lighting fixtures. The following are needed supplies and their costs per unit:
DEMAND
WIRING
(ROLLS)
CONDUITS
WALL
FIXTURES
LIGHTING
FIXTURES
Dormitory
50
100
10
20
Office
70
80
20
30
Apartments
20
50
30
10
PROJECT
ITEM
COST/UNIT ($)
Wiring
1.00
Conduits
2.00
Wall fixtures
3.00
Lighting fixtures
5.00
Use matrix multiplication to compute the cost of materials at each job site.
M5-6
M5-7
matrix R
matrix S
3
2
5
8
0
4
1
2
5
3
2
2
7
1
7
1
2
4
1
2
3
M5-8
6
1
6
3
4
0
6
7
8
9
Find the inverse of original matrix of Problem M5-7 and verify its correctness.
Bibliography
Childress, R. L. Sets, Matrices, and Linear
Programming. Upper Saddle River, NJ:
Prentice Hall, 1974.
M O D U L E
The Binomial
Distribution
LEARNING OBJECTIVES
After completing this module, students will be able to:
1. Describe a bernoulli process.
2. Use the binomial table to solve problems.
3. Apply the binomial probability formula.
MODULE OUTLINE
M6.1 Introduction
M6.2 Solving Problems with the Binomial Formula
M6.3 Solving Problems with Binomial Tables
Discussion Questions and Problems Case Study: WTVX
M6-1
M6-2
Module 6
M6.1 INTRODUCTION
Many business experiments can be characterized by the Bernoulli process, which follows
the binomial probability distribution. In order to be a Bernoulli process, an experiment
must have the following characteristics:
1. Each trial in a Bernoulli process has only two possible outcomeseither yes or no,
success or failure, heads or tails, pass or fail, and so on.
2. Regardless of how many times the experiment is performed, the probability of the
outcome stays the same.
3. The trials are statistically independent.
4. The number of trials is known and is either 1, 2, 3, 4, 5, and so on.
To analyze a Bernoulli process, we need to know the values of: (1) the probability of
success on a single trial, p, and the probability of a failure on a single trial, q (which equals
1 p); (2) the number of successes desired, r; and (3) the number of trials performed, n.
A common example of a Bernoulli process is flipping a coin. If we wish to compute
the probability of getting exactly 4 heads on 5 tosses of a fair coin, the Bernoulli process
parameters are:
p probability of heads .5
q probability of tails (nonheads) 1 p .5
r number of successes desired 4
n number of trials performed 5
There are two ways of solving these Bernoulli problems to find the desired probabilities. The first is to apply the formula, called the binomial probability formula, given in
Equation M6-1.
Probability of r successes in n trials
n!
prqnr
r!(n r)!
(M6-1)
The symbol ! means factorial. To compute 5!, for example, we just multiply 5 4 3
2 1 120. Likewise, 4! 4 3 2 1 24, 1! 1, and 0! 1.
Although Equation M6-1 works well in small problems, it can become cumbersome
when large values of n and r are inserted. The second method is to make use of binomial
distribution tables. Both approaches are illustrated in the following sections.
Using the binomial probability formula, we can solve for the probability of getting exactly
four heads in five tosses of a coin.
p .5
q .5
r4
n5
Probability of r
n!
5!
prqnr
(.5)4(.5)1
successes in n trials
r!(n r)!
4!(5 4)!
54321 4 1
(.5) (.5)
(4 3 2 1)(1)
TABLE M6.1
M6-3
(NUMBER OF HEADS)
PROBABILITY
(r)
5!
(.5)r (.5)5r
r!(5 r )!
.03125
5!
(.5)0(.5)50
0!(5 0)!
.15625
5!
(.5)1(.5)51
1!(5 1)!
.3125
5!
(.5)2(.5)52
2!(5 2)!
.3125
5!
(.5)3(.5)53
3!(5 3)!
.15625
5!
(.5)4(.5)54
4!(5 4)!
.03125
5!
(.5)5(.5)55
5!(5 5)!
or
Probability
120
(.0625)(.5) .15625
(24)(1)
Thus, the probability that 4 tosses out of 5 will land heads up is .15625 or 16 percent.
Using Equation M6-1, it is also possible to determine the entire probability distribution for a binomial experiment. The probability distribution of flipping a fair coin 5 times
is shown in Table M6.1 and then graphed in Figure M6.1.
FIGURE M6.1
Binomial Probability
Distribution When n 5,
p 0.50
Probability of r, P(r)
.4
.3
.2
.1
M6-4
Module 6
MSA Electronics is experimenting with the manufacture of a new type of transistor that is
very difficult to mass-produce at an acceptable quality level. Every hour a supervisor
takes a random sample of 6 transistors produced on the assembly line. The probability
that any one transistor is defective is considered to be .13. MSA wants to know the probability of finding 4 or more defects in the lot sampled.
The elements in this problem would be:
p .13
r 4 defects
n 6 trials
The question posed may be easily answered by using a cumulative binomial distribution table. Such tables can be very lengthy. For the sake of brevity, we present in Table
M6.2 only that portion of a binomial table corresponding to n 6. Other books may contain complete binomial tables for a broad range of n, r, and p values.
Since the probability of MSA finding any one defect is .13, we look through the n
6 table until we find the column where p .13. We then move down that column until we
are opposite the r 4 row. The answer there is found to be .0034, which is a probability
of .0034 that there are 4 or more defects in the sample. This value has been shaded in
Table M6.2.
TABLE M6.2
.01
.02
.03
.04
.05
.06
.07
.08
.09
.10
.0585
.1142
.1670
.2172
.2649
.3101
.3530
.3936
.4321 .4686
.0015
.0057
.0125
.0216
.0328
.0459
.0608
.0773
.0952 .1143
.0002
.0005
.0012
.0022
.0038
.0058
.0085
.0118 .0159
.0001
.0002
.0003
.0005
.0008 .0013
3
4
5
P
.0001
.11
.12
.13
.14
.15
.16
.17
.18
.19
.20
.5030
.5356
.5664
.5954
.6229
.6487
.6731
.6960
.7176 .7397
.1345
.1556
.1776
.2003
.2235
.2472
.2713
.2956
.3201 .3446
.0206
.0261
.0324
.0395
.0473
.0560
.0655
.0759
.0870 .0989
.0018
.0025
.0034
.0045
.0059
.0075
.0094
.0116
.0141 .0170
.0001
.0001
.0002
.0003
.0004
.0005
.0007
.0010
.0013 .0016
.0001
(continued)
TABLE M6.2
(continued)
.21
.22
.23
.24
.25
.26
.27
.28
.29
.30
.7569
.7748
.7916
.8073
.8220
.8358
.8487
.8607
.8719 .8824
.3692
.3937
.4180
.4422
.4661
.4896
.5128
.5356
.5580 .5798
.1115
.1250
.1391
.1539
.1694
.1856
.2023
.2196
.2374 .2557
.0202
.0239
.0280
.0326
.0376
.0431
.0492
.0557
.0628 .0705
.0020
.0025
.0031
.0038
.0046
.0056
.0067
.0079
.0093 .0109
.0001
.0001
.0001
.0002
.0002
.0003
.0004
.0005
.0006 .0007
.31
.32
.33
.34
.35
.36
.37
.38
.8921
.9011
.9095
.9173
.9246
.9313
.9375
.9432
.9485 .9533
.6012
.6220
.6422
.6619
.6809
.6994
.7172
.7343
.7508 .7667
.2744
.2936
.3130
.3328
.3529
.3732
.3937
.4143
.4350 .4557
.0787
.0875
.0969
.1069
.1174
.1286
.1404
.1527
.1657 .1792
.0127
.0148
.0170
.0195
.0223
.0254
.0288
.0325
.0365 .0410
.0009
.0011
.0013
.0015
.0018
.0022
.0026
.0030
.0035 .0041
.41
.42
.43
.44
.45
.46
.47
.48
.9578
.9619
.9657
.9692
.9723
.9752
.9778
.9802
.9824 .9844
.7819
.7965
.8105
.8238
.8364
.8485
.8599
.8707
.8810 .8906
.4764
.4971
.5177
.5382
.5585
.5786
.5985
.6180
.6373 .6563
.1933
.2080
.2232
.2390
.2553
.2721
.2893
.3070
.3252 .3438
.0458
.0510
.0566
.0627
.0692
.0762
.0837
.0917
.1003 .1094
.0048
.0055
.0063
.0073
.0083
.0095
.0108
.0122
.0138 .0156
.39
.40
.49
.50
Source: Reprinted from Robert O. Schlaifer, Introduction to Statistics for Business Decisions, published by McGraw-Hill Book
Company, 1961, by permission of the copyright holder, the President and Fellows of Harvard College.
Expected Value and Variance There is an easy way to compute the expected value and
variance of the binomial distribution. The appropriate equations are:
Expected value np
Variance np(1 p)
(M6-2)
(M6-3)
The expected value and variance for MSA Electronics can be computed as follows:
Expected value np
(6)(.13) .78
Variance np(1 p)
(6)(.13)(1 .13) .6786
M6-5
M6-6
Module 6
Problems
M6-3 This year, Jan Rich, who is ranked number one in womens singles in tennis, and Marie
Wacker, who is ranked number three, will play 4 times. If Marie can beat Jan 3 times,
she will be ranked number one. The two players have played 20 times before, and Jan
has won 15 games. It is expected that this pattern will continue in the future. What is the
probability that Marie will be ranked number one after this year? What is the probability
that Marie will win all 4 games this year against Jan?
M6-4 Over the last two months, the Wilmington Phantoms have been encountering trouble
with one of their star basketball players. During the last 30 games, he has fouled out 15
times. The owner of the basketball team has stated that if this player fouls out 2 times in
their next 5 games, the player will be fined $200. What is the probability that the player
will be fined? What is the probability that the player will foul out of all 5 games? What
is the probability that the player will not foul out of any of the next 5 games?
M6-5 Wisconsin Cheese Processor, Inc., produces equipment that processes cheese products.
Ken Newgren is particularly concerned about a new cheese processor that has been producing defective cheese crocks. The piece of equipment produces 5 cheese crocks during every cycle of the equipment. The probability that any one of the cheese crocks is
defective is .2. Ken would like to determine the probability distribution of defective
cheese crocks from this new piece of equipment. There can be 0, 1, 2, 3, 4, or 5 defective cheese crocks for any cycle of the equipment.
M6-6 Refer to Problem M6-5. Determine the expected value and variance of the distribution
described in Problem M6-3, using Equations M6-2 and M6-3.
M6-7 Natway, a national distribution company of home vacuum cleaners, recommends that its
salespersons make only two calls per day, one in the morning and one in the afternoon.
Twenty-five percent of the time a sales call will result in a sale, and the profit from each
sale is $125.
(a) Develop the probability distribution for sales during a five-day week.
(b) Determine the mean and variance of this distribution.
(c) What is the expected weekly profit for a salesperson?
Case Study
WTVX
WTVX, Channel 6, is located in Eugene, Oregon, home of the
University of Oregons football team. The station was owned
and operated by George Wilcox, a former Duck (University of
Oregon football player). Although there were other television
stations in Eugene, WTVX was the only station that had a weatherperson who was a member of the American Meteorological
Society (AMS). Every night, Joe Hummel would be introduced
as the only weatherperson in Eugene who was a member of the
AMS. This was Georges idea, and he believed that this gave his
station the mark of quality and helped with market share.
Case Study
M6-7