THE UNIVERSITY OF TENNESSEE AT MARTIN
College of Business and Public Affairs

Subject: Statistical Analysis
Course Number: Management 720
Course Title: Statistical Tools of Quantitative Analysis
Credit Hours:  3
Contact Hours:  3 hours/week
Course Pre-requisites:  Graduate Standing
Instructor:  Dr. John Knight
Office Location: Business Administration Bldg. 229A
e-Mail Address: jknight@UTM.edu
Office Hours:  TTh 2:15-5:15  TTh 5:50 - 6:50 (Tuesday in Martin, Thursday in Jackson), TTh 9:25 - 10:25 P.M. (Tuesday in Martin, Thursday in Jackson)
Office Telephone: (731)587-7351
Course Text and Related Material:
Using Statistics to Analyze Business Problems: Case Studies with Computer Applications by John E. Knight and Advanced Statistical Decision Making with Cases by John E. Knight  (Available in hard copy and distributed in class.)  Additional References:  Any good library book on statistics will be a good reference or use the http://www.statistics.com/  link for a series of other on-line textbooks (see StatSoft link http://www.statsoftinc.com/textbook/stathome.html for an especially good text).

Disabled Students:  Any student eligible for and requestiong academic accommodations due to a disability is requested to provide a letter of accomodation from P.A.C.E. or Student Academic Support Center within the first two weeks of the semester.

COURSE PURPOSE, GOALS AND OBJECTIVES:
Course Description: Designed to give students the statistical skills necessary for advanced work inthe functional areas of business administration.

General Course Objectives:
(1) To present the basic univariate and multivariate statistical methods utilized in business decision making in a case oriented teaching method.
(2) To develop the student's ability to analyze statistical data utilizing computer analysis.
(3) To develop a student's ability to perceive a statistical situation, collect data, analyze data, and reach conclusions that can be related to managers.
(4) To encourage student participation in the discussion of cases and other reading materials.
(5) To engage the students in the practical and real life discussion of advanced statistical studies.
(6) To gain an awareness of the use of statistical methods in business literature.

Expected Student Outcomes:
(1) Students will be able to interact with web-based material including downloading of programs and files.
(2) Students will learn a standardized statistical program for the analysis of statistical data.
(3) Students will learn how to determine if data are valid, how to search a database for porblems, how to analyze single variables in a database, how to analyze relationships between two variagbles in a database and how to analyze and interpret three or more variagbles and relationships in a database.
(4) Students will learn to independently analyze databases and make statistical summaries of their findings.
(5) Students will become familiar with readable statistical literature that provides a practical basis for the use of statistics in managerial roles.
(6)  Students will become familiar with statistical design of experiments and multivariate testing and the statistical process of improvement that it culminates.

Course Outline: Details of each session will follow the general course outline given below.  The actual dates and topic presentations may vary slightly due to a multitude of factors.  Due to potential sickness or other University assignments,  the session topics may be slightly modified as the semester progresses.
 

Sessions 1-5 Analyzing Single Variables: Summary Statistics, Confidence Intervals, Sampling Error, Hypothesis Testing and Simulations: Presto Pizza Company.  Analyzing Two Variables at a Time: Regression, Correlations, Analysis of Variance and Crosstabs: Metro Realty Company
Sessions 6 Test #1 Diagnostic Test
Session 7-11 Advanced Case Studies:
     Marketing Applications:  Cycle World
     Management Application of Wage Discrimination: Service Pro
     Management Application of Employee Ratings:  Fabor Manufacturing Company
Interaction and Curved Line Relationships:
     Management Application on Work Standards:  Green Seed Company
     Management Application of Productivity Improvement:  Universal Manufacturing Company
Sessions 12-15 The Statistical Improvement Process: 
     Steps 1-3:  Creating the Environment, Key Variables and Operational Definitions
     Steps 4-6:  Statistical Process Control:  DataStor Company
     Step 7:  Design of Experiments: 
          Screening Experiments: Airplane Example 
          Refining Experiments and Full Factorial Designs:  Checker's Pizza
Final Exam
 

COURSE REQUIREMENTS/EXPECTATIONS:
Grading Procedure:  Grades will be based on the percentage of total points that are possible during the course.
         A    87.5% and above
         B    75% and up to 87.5%
         C    62.5% and up to 75%
         D    50% and up to 62.5%
         F    below 50%

Class Policies:
Class Attendance Policy:  Class will be taken each day precisely at the scheduled starting time based on the computer time.

Testing and Other Procedures:  In order to insure student equity,  the University Office of Equity and Diversity has approved the following procedures regarding testing.  All tests will be passed out face down one at a time in sequential order.  Once all of the test papers are in front of the students and the instructor gives permission, the students may turn over the test. When the time for the test is over (based on the computer time), all remaining students will be required to immediately cease writing and stand behind their desk.  The students will remain standing until I personally collect each paper at which time they will be dismissed. Any student refusing to stand and stop writing at the designated ending time will receive a grade of zero.  Any student looking at the test before permission is given will receive a grade of zero.  Any student that keeps their hands on the test and not surrendering the test will be given a grade of zero.  The instructor has the right to call Safety and Security to collect the test from any student that will not surrender the test to the instructor when instructed to do so.  The test is considered University property and the studentís comments on the paper do not make the paper their property.  Thus, the instructor retains to right to collect each and every test that has been distributed to the students.   Although the test paper is considered University property, the tests will be destroyed 2 weeks after the beginning of the next semester (including summer) as per general practices established by University tradition.

In an additional effort to dispel any perceptions of discrimination, the instructor will not be give ANY ASSISTANCE OR ANSWER ANY QUESTIONS DURING THE TEST.  You must work the test completely on your own just as every other student.  Additionally,  on any test problem that has a specific numerical answer,  either the answer is correct or incorrect.  No part credit will be given.  If an error in a preceding portion of the problem affects the answer in the next sub-problem,  only the correct answer will receive credit.  Questions requiring written answers will obviously receive variable weights from the instructor.

Finally, all student questions must be asked and posed before the FINAL CLASS PERIOD.  Office hours are officially over on the last day of class.  There will be no special study sessions prior to test dates as these sessions do not allow all students equal chance to attend given work and other commitments.

Academic Honesty:  All students are expected to do their own test work.  Any student found to be cheating will be handled according to the regulations in the University Catalog.

An On Line Link to Statistics that you may find of value is http://www.statistics.com/. This link provides you with access to the statistical world.  You may want to explore that world when you have time. In fact,  if you happen to be somewhere and don't have access to the statistical program that is used in this class,  you can consult this page and consult "statistical calculations right on your web browser" and have most any calculations performed on line.
 
 

Case #1:  Analyzing Single Variables: Summary Statistics, Confidence Intervals, Sampling Error, Hypothesis Testing and Simulations

PRESTO PIZZA COMPANY

In Case #1- #2,  the basics of statistics from data classification through hypothesis testing is presented in the context of a case.  The student will be responsible for reading and understanding the text material associated with the case.

The specific problem is given in the text and can also be found on line under the following link:  Introduction and Presto Pizza.
The presentation of the material will focus on the specific case with the background statistical knowledge and explanation of techniques given in the text. The remainder of the background knowledge necessary for the Presto Case is given in the following link:  Background Material (2).

Please note the following.  The notes that are herein linked to the web page are copyrighted. At this time,  the notes are for use only by my students. They can be downloaded by the students for their direct use but not reproduced commercially or for other purposes.  However, you will need a User Name and Password as the files are for use by registered University of Tennessee students only.  These access codes will be provided upon registration in the class.
 

The computer program that will be used to analyze statistical data is A Statistical Program (ASPW)  (see www.dmcsoftware.com). The program is available on the student computers on campus including the School of Business Lab in Room 27 and in the computers in the Humanities Building (IBM computers).  However,  if you have your own computer or would like to download to your own computer, you can click on the ASPW link.  When you click for the link,  you will be asked where the file should be saved.  You can either save it to a disk in your a:/ drive or you can install it into your permanent C:/ drive.  The program will guide you through the set up.   The School of Business Administration at UTM has obtained a site license that allows all UTM students to download the program onto their computers for use by the student.  Another option is to download the program to your DESKTOP and install on your computer by clicking on the Setup icon and following the installation instructions.  The ASPW program contains an alpha-numeric graphic program that is quick and efficient.  For more exotic data presentations,  an additional program called ASPG (A Statistical Program Graphics) is attached and can be downloaded to your DESKTOP and installed on your computer.  Please install these programs on your computer.

A computer program such as ASPW (or any other program) executes mathematical operations on data.  Data however must be collected and supplied to the computer program.  Many data files might be utilized in the class.  To begin, please download the computer file "pron386.sf" from the data link that follows.  The data files are already formatted for use in ASPW.  If you
require the files in ASCII format for downloading to another statistical program,  please let the instructor know.   The data link for files is Data Files  located on the UTM server.  For the Presto case,  the file PREST386.SF will be utilized for analysis.

One important concluding concept makes the Presto Pizza Case extremely interesting.  Since statistical data is normally kept in spreadsheet format,  the manipulation of those columns of data can often allow us to investigate "what if" scenarios and evaluate some possible problem alternatives.  For example, based on the statistical and judgmental analysis of the Presto Pizza Company, far more than 5% late pizzas can be expected.  Will our recommendation be to declare bankruptcy immediately or can we use the data to test some "what if" possibilities and provide Tony with some options that he may implement? For example,  what about a 30 minute guarantee,  or what about a new oven or a more efficient rearrangement of the facilities, or what about restricting the delivery radius that is guaranteed?  These and other alternatives can be assessed for their impact on the likely percentage of late pizzas.

At this point,  try to list a minimum of 25 ideas that might directly help reduce the percentage of late pizzas.  The concept of idea generation and decision making will be explored in relationship to generating ideas for solving the Pizza problem.  One of the best texts that I know that gives a clear indication of the relationship between generating ideas and decision making is the New Rational Manager by Charles Kepner. Without ideas and their economic analysis,  efficient utilization of financial assets will not be achieved as the incremental rate of return will not be properly integrated into the decision making sequence.

An article of interest will be important to read as background for Sessions #1 - #2.  The article is linked below in pdf format.  You may have to download Adobe Acrobat Reader in order to view the file. "Assumptions for Statistical Inference" by Gerald Hahn and William Meeker, The American Statistician, Feb. 1993, Vol. 47, Issue 1, page. 1.

An internal class communication systems will be provided to the class through the internet.  The link will be http://blackboard.utm.edu.  You should be automatically registered in the class and be able to communicate with other members in the class via the communications link.  Blackboard will also provide a method for discussion boards.  Please make use of these communication tools as they are presented to the class.

I will expect you to have read the Hahn and Meeker article before the midterm class period.  Once you have completed the reading,  please use the communication system and the discussion board to discuss the article.   If you are interested in contacting a particular person in the class, e-mail them directly to their personal e-mail account.  Likewise,  if your question is to be addressed directly to me and not to be shared by others, use my e-mail account at jknight@utm.edu.  Typical items to post to the discussion board or to e-mail are:
    1.  Comments on articles that you have read.
    2.  Questions to the class about how to work a problem.  (I prefer to let the students attempt to answer the questions.  I will monitor the responses and step in with a correction if need be.)
    3.  Sharing of information on library searches, etc.

Once the Presto Pizza case has been fully explored,  writing up the results into a convincing and succinct report is crucial.  In an effort to provide guidance into the manner in which I would expect the report to be written,  I have attached a short report on Presto Pizza that you can review and consider at the end of the questions to guide your study.  Note that this Summary is written so that a statistical layman or manager could have reasonable understanding of the issues while those with greater statistical training could also be appraised of the inferential properties of the tests and their confidence.  Note that the summary would be backed up with many potential appendices and other analyses of less importance.  These reports and analyses have not been mentioned in the summary report but would be available upon request (or at a computer terminal if the analyst were so asked).  I will expect discussion questions on exams to follow the format given in the summary report.

HOMEWORK EXPECTATIONS:
    1.  Read and Review the background text material for the Presto Pizza case.  These notes review the basics of statistics and provide the foundation for the study of further statistics.  You will be expected to answer general questions about the material on the first exam.  Some potential discussion questions from the material would include.
    (a)  Discuss the meaning of each term in my given definition of statistics.
    (b)  Is the job of a manager to MOTIVATE all the employees to be "above average"?
    (c)  Sketch out a normal distribution with a mean of 55 and a standard deviation of 3.  What percent of the distribution is greater than 55?  What percent of the distribution is greater than 56.3?  etc.
    (d)  Discuss the concept of sampling error and the standard error of the mean.  Discuss the two (or three) potential approaches for reducing or controlling the sampling error.
    (e)  Discuss the notion of an enumerative versus an analytical population and the impact that the type population will have on the computer generated numbers from statistical analysis.
    (f)  Discuss the impact of the terms of randomness and representativeness on the sampling process and the impact that these have on the computer generated numbers from statistical analysis.
    (g)  Discuss the "conclusion" of an hypothesis test - especially in regards to what is concluded and what is not with any given confidence.
    (h)  Discuss the terms "statistically significant differences" versus "practical differences" in regards to statistical analysis.
    (i)  Discuss the difference between "independent populations" and "matched pairs analysis".

        These questions should be answered and understood before the Midterm Test.  These questions will compose about 20% of the first test and I have noticed that many of the points missed by students results from poor answers to these questions.

2.   Install and get the ASPW program (and graphics) working on your computer.   Attempt to download and work with the PREST386.sf file.

3.  Complete all of the questions with appropriate answers in the Presto Case.  Attempt to complete before Class Period 3.
 


Case #2.  Analyzing Two Variables at a Time: Regression, Correlation, Analysis of Variance and Crosstabs

Metro Realty Company

Metro Realty faces a crisis.  Two of the agency's former customers have recently made a charge to the Board of Realtors that the agency didn't represent their best interests.  In fact,  they felt that they had been encouraged to sell at substantially below market prices.  A recent article in the newspaper not only sparked their interest but also gave them evidence to potentially support their claim.  The agency had to dispel this allegation in order to continue to be an viable growing agency in their geographical area of concentration.  Your job is to assist James Gleason in analyzing his data in an effort to clarify the allegations and charges.   The questions for the case are found in the Metro Realty link.  The supporting text on the appropriate statistical technieques are given in the Two Variables at at Time link.  The appropriate data file is METRO.SF.

Techniques that emphasize the analysis of two variables at a time will be emphasized in this case:  regression analysis, correlation analysis, analysis of variace and crosstabs.  However,  questions that review the one variable analytical techniques will be reviewed and also prove important in this case.  Finally,  a true multivariate techniques that handles 3 or more variables at a time will conclude the case and provide important insight into the problems that the company faces.

HOMEWORK EXPECTATIONS:
    1.  Continue to read articles and develop answers for the questions suggested for the first test.
    2.  Download the METRO.sf file and work all of the problems provided in the Metro Realty link.
 


First Test - 3 hours

The Midterm exam is primarily designed to encourage the students to be prepared to demonstrate competency in the basic techniques covered to date and to be able to utilize the computer in making the appropriate statistical calculations.  The test will cover the two cases presented to date, Presto Pizza and Metro Realty.  Given that some of the material is still new in your mind,  I have provided a Review Sheet that can be utilized on the test.  The review sheet should help you in formulating the technique to use and the method for establishing the hypotheses and conclusions. Obviously, you will also need to be familiar with the computer usage.

Additionally,  several discussion questions will be given based on the articles assigned to date and on any related topics discussed in class.


Cases # 3 and 4.  Cycle World and Service Pro

Once the diagnostic test has been completed, you should have the background to work some problems more independently.




CYCLE WORLD, INC.

Cycle World is a regional sport cycling company headquartered in Seattle but with 5 other branches in the Northwest.  They are interested in hiring a marketing agency,  Keels, Inc.,  to study their customer profiles and decide the type of national magazine advertising that would be most beneficial.  The specific information regarding the case is given in the text material line on Additional Cases.  The computer file being utilized is keels.sf.

The Keels Agency sampled 300 customer profiles from the existing customer base to find out more about customers that presently are attracted to Cycle World business.  Knowing this information,  Keels hopes to recommend to Cycle World the best way to spend a projected advertising budget of $240,000 to advertise in national magazines and to expand their market nationwide.    Each magazine selected results in an expenditure of about $80,000 when selected.  Of course,  Keels also must be paid.  Obviously, to earn their fees,  Keels should be able to increase the effectiveness of the selection process so that a reduced amount of advertising will be more effective and still allow them to earn their fees.

SERVICE PRO, INC.

Service Pro, Inc. handles customer requests from telephone callers.  Three classifications of agents employeed, from new hirees to the most experienced represenatives.  A couple of years ago,  Service Pro decided to implement an affirmative action program in order to increase the percentages of minorities in its employment.  The company felt that the program had been successful.  However,  recently two minority employees have approached top management regarding alledged salary discrimination given that the minorities on the average are paid far less than the other employees.  Management has gathered data on the employees and has requested a full statistical study to determine if salary discrimination in any form exists.The specific information regarding the case is given in the text material line on Additional Cases.  The computer file being utilized is servpro.sf.

Complete the two cases above and be ready to discuss during the next class period.  I have developed a series of questions for you to review that will guide you through the logic of the cases.  Please attempt to confirm answers for all of the parts of each case and please come prepared to discuss each case.


Case #5. Further Case Studies:
          Management Application of Employee Ratings:  Fabor Manufacturing  Company

Cycle World and Service Pro will be discussed and the appropriate solutions and answers presented.  The first test will then be returned with time for the students to review their performance on the test. Finally, the Fabor Manufacturing case will be presented in class and the student will complete the remainder of the case on their own time.
 
 

FABOR MANUFACTURING COMPANY

The top management at Fabor Manufacturing Co. has decided that a reward system should be developed to motivate employees.  In fact,  a bonus pool of money has been established to be divided up among the top 10% of the employees.  In order to identify these employees,  the company has decided to use last year's supervisor's ratings of the employees.

Numerous strategies have been suggested to find the best 10% of the employees but some real problems seem apparent as the actual implementation approaches.

As an added portion of the case,  the data will be transferred to as ASCII format and them exported to a standard spreadsheet appliciation (Microsoft Excel) where one particular aspect of the analysis might be more easily handled.  This exercise will also illustrate how data can be imported and exported to the ASP Statistical package.  This procedure might be especially useful since many databases are maintained in standard spreadsheet form.

The data file for the case can be found in the data link under glenco.sf.  The list of directed questions can be found in the EmployeeBonusProgram link.

The conclusion that might be drawn by completing the analysis of the cases can be summarized as follows. The rigorous and exhaustive application of the simple techniques covered to date will provide most managers with 80% of the technical information in the data.  Obviously more sophisticated analysis by even more advanced techniques (see www.statistics.com link with "on line textbooks" and "Electronic Statistics Textbook" by StatSoft) can possibly find a few more gems in the data.  However, the Pareto principle probably applies at this point in that the methods shown to date can usually garner 80% of the information using only 20% of the techniques available


Cases #6 and 7.  Interaction and Curved Line Relationships:
     Management Application on Work Standards:  Green Seed Company
     Management Application of Productivity Improvement:  Universal  Manufacturing Company

Simple regression and correlation assumes that a scatter plot of the data will result in a general straight line to best describe the pattern in the data points.  However,  in some situations,  the scatter plot of the data will indicate a curved line relationship exists in the data.  Many examples of curved lined relationships exist (although the simple regression assumption is the most prevalent description of data).  For example,  if one plotted the number of cases of soft drinks sold (y axis) versus the average monthly temperature,  one will find that the sale of soft drinks increases at an increasing rate as the temperature increases - a curved line relationship.  Additionally,  if one plotted the orders were week placed at a manufacturer (y axis) versus the average delivery delay quoted (quoted time to fill the order),  the curve would asymptotically approach both the x and y axes.The concept of linear transformations in data will be used to attempt to describe asymptotic curves while 2nd and 3 degree power curves will describe other curved line relationships.  Some examples will be given.

In model building and later is Design of Experiments,  the concept of interaction is very important.  Interaction implies that the incremental effect of an independent variable y on a dependent variable x will be different depending on the presence of a second independent variable.   Thus,  in the Easton Realty case,  the incremental increase in price per square foot of a house may in fact be different in Dallas than in Fort Worth.  In a factory situation,  the final tensile strength of a piece of metal may increase dependent on the amount of carbon in the metal but the increase may be at different incremental rates in the presence of different annealing temperatures.  The concept of interaction and its understand will be a new major emphasis of statistical understanding and application in the near future.

The statistical background and foundational information on these two topics is given in the Curved Line and Interaction link.  Please read this information before working the following two cases.

GREEN SEED COMPANY

The Green Seed Company takes bulk seed and packages the seed into individual seed packets for the home gardener.  The 200 varieties of seed are then placed in a runway in alphabetical order where the seeds are then "picked" by order pickers and placed into cardboard display units such as one finds in stores such as Walmart, Lowe's, etc.  Order picking is a very common operation in warehousing operations.  Thus, this case serves as a practical example of the use of statistics.  The data is actual data derived from time study.

The objective was to develop a predictive time to pick each individual reorder (or mail) based on the number of units ordered and on the number of different variety of seeds to be ordered.  The employees were paid on piece rate and so the issue of time was important to them.  The company likewise had to pay the salaries so time translated to money and that was important to them.  The issue was "what is a fair time" on any one order ... when all the orders are different?  Can that be predicted?  If so,  then many issues would be settled including being able to schedule jobs and employee levels more accurately.

A sample of 29 different orders were constructed as a representative sample of different "reorder sizes" that were being processed. Each order was timed by an industrial engineer to determine the normal picking time per unit.  Two different pickers were  times to determine if the engineerings leveling factors were consistent.  Data was collected on the normal time/unit,  the number of units picked, and the number of locations (varieties) to be picked.

The pickers were quite aware of what was called "good orders" and "bad orders".  In order to make a good piece rate pay,  competition was fierce over who got the good orders.  A good order was one that had a large number of units with a small number of locations since units could be picked and counted much faster than locations could be found!   A solution was needed that would also integrate into the company's computer system (COBOL).

The case file associated with this case is "mailpck.sf" in the Data Files link.  The case will be presented and solved in class and will utilize the concepts of curved lines and interaction to find a variety of different potential equations to describe the relationships.

UNIVERSAL MANUFACTURING COMPANY

Universal Manufacturing Company has three different plant locations.  The home office has been interested in increasing the productivity of the company as a whole.  In fact the company has even toyed with the idea of introducing "supervisory" performance ratings based on the "productivity" that each plant achieves.  When confronted with these proposals,  the supervisors raised some legitimate concerns about which orders would be sent to which plants ... as they felt the assignment of different orders to different plants had more to do with productivity than their own personal efforts or that of their employees.  Data was collected from the three different plants on the size of the orders assigned and the productivity that was achieved on each order.  You have been assigned as the statistical analyst to answer questions that arise over the disputes with using the new performance system.  A complete discussion of the case (questions and answers) found in the Productivity link.  The related computer file is univmfg.sf.

Using the material in the material in the Curved Line and Interaction link and the experience of analyzing the Green Seed Company,  attempt to generate the same answers that are given in the problem.  Although such a problem seems difficult and obtruse without assistance,  study of the problem and answers will greatly enhance your understanding of how the questions posed by the supervisors can be viewed in statistical terms and how the analysis by the statistician can lead to a result that is acceptable to the supervisors and that will truly improve the productivity of the overall company.
 


Case #8. Improving Business Processes using Statistical Methods
     Step 1. Creating the Environment
     Step 2. Key Variables and Operational Definitions
     Step 3. Creating a Capable and Controlled Measuring System
     Steps 4-6:  Statistical Process Control:  DataStor Company

The first portion of the course covered the basic techniques of analyzing data.  While analyzing the data,  comments were continually made in reference to the key variables, difficulties with data,  how poor planning diminished the value of the data, and how the rigorous analysis of poor data would still lead to poor results (garbage in - garbage out).

Oftentimes,  individuals knowingly or unknowingly collect and/or present information poorly and mislead unsuspecting individuals.  One interesting supplemental book to read is "How to Lie with Statistics" by Huff and Geis.  Although this course attempts to teach the correct presentation of data,  one should always be aware that others can utilize numbers to create a false presentation in order to sway one to erroneously agree with the presenter.   At some future time, you may want to take time to read this short but famous book. At this point, it is not required for class.

The article Data Collection and Analysis:  Some Managerial Issues by Ed Knight should be reviewed as background information to this session.   These comments will serve as as a platform from which the remainder of the course will be launched.  Emphasis will be placed on PLANNING AND COLLECTING  THE APPROPRIATE VALID AND RELIABLE DATA.  Once that is accomplished,  the analysis of data will be more effective - whether using the techniques presented to date, material yet to be presented in this course, or additional more sophisticated techniques in other references.

The ultimate result of all of our efforts in statistics should be in using the techniques appropriate to improve business processes.  The remainder of the course will focus on the correct procedures for DRAMATICALLY IMPROVING BUSINESS PROCESSES using statitics.  The correct procedures will be explained as a series of 7 steps discussed below.

In Step 1,  the concept of creating the environment for improvement is emphasized including (a) identifying a significant problem, (b) forming an improvement team and (c) finding an individual able to break down improvement barriers.

In Step 2,  the concept of "key variables and operational definitions" are developed.  As a humorous preview of the material you'll examine in this chapter,  you may want to view the Lunch time picture and remember its implications as you discuss the topics, especially in regards to Operational Definitions. A company that sets a limit on noontime drinks at a business meal may need to re-evaluate its operational definition.

In Step 3,  the development of a capable and controlled measurement system is discussed.  Terms like "accuracy", "precision", "capability", "controllablility" and "repeatbility and reproducibility" will be presented and discussed based on statistical theory.

Steps 4-6 focus on the basic concepts Statistical Process Control. Although Dr. Walter Shewhart was known as the original developer of SPC,  Dr. W. Edwards Deming is by far the most famous. Although a consummate statistician,  he may well be finally known for his MANAGEMENT THEORIES most comprehensively published in his book "Out of the Crisis".  You might also want to do a world wide web search using the term W. Edwards Deming to study more about this remarkable man.

Dr. Deming was well known in Japan having helped the Japanese out of the industrial decimation of World War II.  Until 1980, he was relatively unknown in America until he was accidently "discovered" at the age of 80 through an NBC News documentary titled "If Japan Can,  Why Can't We" (see article and links below).  From that telecast until his death at the age of 93,  he lectured and cajoled most every major CEO in the United States to consider his system of management that was based on "statistical logic and foundations".

Dr. Demings' theories as based on the premise that VARIATION IS THE ENEMY of quality.  The reduction of variation will lead to better quality, service, and more business.  The age of customer service and product quality can be directly traced to Dr. Deming's impact on the world.  Many quality games can be used to demonstrate how a basic philosophy in statistics can dramatically affect the way in which a manager conceptualizes the managing process. An explanation of the "funnel experiment" is given in the reference and link below.

A particularly interesting article about Dr. Deming relates to his "discovery" as a guru at the ripe age of 80 ("Discovery of the Prophet of Quality" by Clare Mason).  The famous funnel experiment that Dr. Deming illustrated to show how poor statistical logic affects managers is found in the linked articles.

Finally, two very good articles summarize SPC and the background and theory: "A Graphical Exploration of SPC" by Robert Hoyer and Wayne Ellis,  Quality Progress, May 1996  (Part 1) and  "A Graphical Exploration of SPC" (Part 2), Quality Progress, June 1996.  One program emerging from Demings conversion of management to a quality philosophy is the Six Sigma methodology of Motorola.  A recent article, "Implementing the Six Sigma Solution" by Jerome A. Blakeslee, Jr.,  Quality Progress, July 1999, p. 77-85 will provide one with the basic background of the methodology. Six Sigma is a trademark of Motorola, Inc. and is taught by Advanced Systems Consultants. You can obtain more free articles on Six Sigma from the link provided.

The primary statistical information for this class is found in Class Session 18.  SPC.
 

HOMEWORK EXPECTATIONS:

1.  Read the following articles.
 (a)  "Discovery of the Prophet of Quality" by
 (b)   Article on funnel Experiment.
 (c) "A Graphical Exploration of SPC" by Robert Hoyer and Wayne Ellis,  Quality Progress, May 1996  (Part 1)
 (d)  "A Graphical Exploration of SPC" (Part 2), Quality Progress, June 1996.
 (e) One program emerging from Demings conversion of management to a quality philosophy is the Six Sigma methodology of Motorola.  A recent article, "Implementing the Six Sigma Solution" by Jerome A. Blakeslee, Jr.,  Quality Progress, July 1999, p. 77-85 will provide one with the basic background of the methodology. Six Sigma is a trademark of Motorola, Inc. and is taught by Advanced Systems Consultants. You can obtain more free articles on Six Sigma from the link provided.
2.  Complete the analysis of the DataStor Company.

Homework Expectations:

1. Determine a process that you would like to improve and apply the first three steps of the improvement process to the system.  Give special emphasis to the selection of the key measure,  how you will define it,  how you will measure it,  how you will prove what you measured is repeatable, etc. Some examples of improvement targets are suggested below.
    (a)  the quality of undergraduate education at a University.
    (b)  the percent of customer complaints.
    (c)  the percent of product with flaws
    (d)  the percent of students passing an exam
    (e)  the product yield from a given raw material
    (f)  the time that a battery will operate a CD player
 Obviously, many other improvement targets exist.  Try to select one that would be of interest to you based on your experience and background.
 
 

DATASTOR COMPANY












Class Period #8.  Step 7:  Design of Experiments:  Checker's Pizza

 

Design of Experiments (or Multivariable Testing) has gained increasing popularity in the past years.  In addition to the readings given below,  the technical information on the background for DOE and some calculation routines are provided in Session 19. Design of Experiments.

Originally,  DOE was developed under the concept of Analysis of Variance and especially multifactor ANOVA.  Using multiple factor ANOVA is often times easy and beneficial but more complicated designs require advanced training and more sophisticated planning, design, and calculations.  The concepts of Sum of Squares is extensively utilized in the calculation of ANOVA and multifactor testing when using this concept.

The method that I will be presenting for multivariate testing is much easier and will handle far more multiple factors while making some limiting but efficient assumptions.  I will introduce the Plackett-Burman DOE's as examples of how to use DOE in business and industry.

1.  Read the following articles.
    (a)  "The New Manta:MVT" by Rita Koslka, Forbes Magazine, March 11, 1996.
    (b) "Statistics drive Lowe's marketing initiatives",  National Home Center News,  May 03, 1999 which can be found under the Library link of BizSuite using the search term "multivariable testing".
    (c)  "BellSouth Uses Scientific Testing to Increase Call Center Sales"  by Curren, Dariel,  TeleProfessional,  June 1998, which also can be found under BizSuite.
2.  Complete the analysis of the case Checker's Pizza from Session 20.
3.  Complete the homework problem cases in Session #19 using full factorial methods and using the Plackett Burman designs.

Session #19. Design of Experiments
 

4.  Design a Plackett Burman screening experiment that will determine the important factors in helping people to lose weight.
 
 

Class Period #15.  FINAL EXAM

      The final exam will be composed of the following portions with associated weights.
        (a) Case Analysis - 70% - One or two cases will be given for in class analysis. Specific questions will be asked similar to the midterm exam on Pronto and Easton.  At least one of the cases will be completely new material while the other will be one that we have already analyzed.
       (b)  Summary of Case - 10% -  You will be asked to summarize a solution for one of the two cases given in part (a).  The summary should include your rationale for solving the problem AND provide salient statistical information which indicates why you chose the solution you presented.
        (c)  Discussion questions - 20% -  Two discussion questions will be asked from the following list of questions.
    DISCUSSION QUESTIONS
            (1)  Discuss three reasons why Design of Experiments is superior to One-Factor-At-a-Time testing in process improvement activities.
            (2)  Describe the term interaction, draw graphs of that it looks like,  and tell how interaction is found using the original data and the computer program.
            (3)  Describe the process of finding the "best" equation when a scatter plot shows that the data in non-linear (that is, a curved line).
            (4)  Describe the rationale for each step of he 7 step improvement process.
            (5)  Describe typical "survey" topics and the basic elements of the topics, e.g. Attitudes (with subpoints), etc.