Chairman's Message

r

by Ed Mykytka

As May draws near, the Statistics Division (SD) is actively preparing for its participation in the 41st Annual Quality Congress, to be held this year in Minneapolis, Minne-sota on May 4

fi"4, through May 6. Our program chair-man, Bob Perry, has lined up a number of exciting presentations within four Division-sponsored technical paper sessions.

Two of these sessions will be high-lighted by tutorials based on recent volumes in the "How-to" booklet series. Neil D. Cox will present mate-rial from his booklet entitled "How to Perform Statistical Tolerance Analysis" and Thomas P McWilliams will speak on "Sequential Methods in Quality Control." We are also trying to arrange informal question and answer sessions with both authors at the Congress. Times and locations will be announced at the sessions and posted at the Statistics Division booth.

Another innovative feature will be the presentation of case studies in quality and productivity by Lynne Hare and Blan Godfrey. Both presenta-tions are offshoots of last year's very successful panel discussion on "The Statistician's Role in Quality Manage-ment" and should be well worth hear-ing.

We are also pleased to be supported by other fine authors and speakers, including Gordon Constable, George

Tagaras, Hua Lee, James Kenney, Louis Cabana, Tom Barker, Madhav Phadke, and Carey Parks.

We shall very soon be finalizing plans for the 1988 Annual Quality Congress to be held in Dallas. If you have any suggestions for the program, or would like to submit an abstract, please contact our program representa-tive, George Marrah, at (703) 568-6534.

Fall Technical Conference

Conference-wise, the Division is enjoying another very good year. The Fall Technical Conference (FTC), cos-ponsored by the Chemical and Process Industries Division (CPID), and the Section on Physical and Engineering Sciences of the American Statistical Association (ASA), was a tremendous success. Attendance reached an all time high and the quality of the confer-ence was superb. Congratulations are extended to Jim White, chairman of CPID, and the program committee, headed by Dave Sylwestre, for an excellent conference.

The preconference courses spon-sored by the Division were attended by overflow crowds and received excellent reviews. Thanks go to Dan Boroto, Andrea Zahn, Doug Zahn, and Dick Trout for their hard work and outstanding contribution to the pro-gram.

Plans are well underway for the 1987 FTC in Atlantic City on October 22 and 23, with preconference courses on October 21. Since there is no local

ASQC section in Atlantic City, volun-teers would still be very much appreci-ated to help out on many levels. If you would like to lend a hand, particularly if you reside in the nearby region, please call Frank Sinibaldi, chair-elect of CP1D, at (914) 578-7107 or Roger Hoerl, SD program representative, at (215) 522-5011.

Values Workshop

Our most unique activity this year was a workshop entitled "Using Values to Improve the Effectiveness of Statisti-cians" which preceded the ASA Winter Meeting in Orlando, Florida. The brainchild of our Division treasurer, Lynne Hare, the workshop was led by Dr. Charles Dwyer of the Wharton School. Details about the workshop are available elsewhere in this Newslet-ter.

I simply would like to thank Lynne for his hard work at organizing the workshop and recognize Blan Godfrey of AT&T Bell Labs, George Dorman, vice-president for Corporate Quality at Westinghouse, and Dave Fluharty of the Ford Motor Company for their valuable contributions.

By the way, Lynne was recently elected to the Executive Board of the Quality and Productivity Committee of ASA, ensuring that our two organiza-tions will continue to enjoy good working relations. (We were also pleased to have that committee as a cosponsor of this workshop.) Congrat-ulations, Lynne.

Continued on Page 4

ANNUAL QUALITY CONGRESS MAY 4-6, 1987 HYATT REGENCY MINNEAPOLIS, MN

In Memoriam William G. Hunter-1937-1986

was the author of over 80 technical articles, as well as co-author (with George Box and Stu Hunter) of the textbook Statistics for Experi-menters.- An Introduction to Design, Data Analysis, and Model Building.

Besides being founding chairman of the ASQC Statistics Division in 1979-80, he was chairman of the ASNs Section on Physical and Engi-neering Sciences in 1981, and he had also been active over the years with the ASQC Chemical Division (now called the Chemical and Proc-ess Industries Division). It is thus appropriate that the next Fall Tech-nical Conference, jointly sponsored by those three organizations (and to be held October 22-23, 1987 in Atlantic City), has been dedicated to Bill's memory.

The title of the Youden Address that Bill gave at the 1983 Fall Techni-cal Conference—"Learning about the world around us by using statis-tics as an aid for listening to and conversing with it" —tells a lot about the way in which Bill approached scientific investigation in particular as well as life in gen-eral. Bill's goals were excellence in teaching applied statistics and pro-moting improved quality and pro-ductivity in American industry and government.

Bill will be missed by his peers, both as a colleague and as a friend.

Dr. William G. Hunter, founding chairman of the ASQC Statistics Division, died of cancer on Decem-ber 29, 1986 at his home in Madi-son, Wisconsin at the age of 49.

Bill graduated from Princeton in 1959 and went on to earn a master's degree in chemical engineering from the University of Illinois in 1960, as well as a master's degree and a doctorate in statistics from University of Wisconsin-Madison in 1962 and 1963, respectively.

Bill was a Professor of Statistics and Industrial Engineering at UW-Madison. In 1985, Bill became the founding chairman of the Center for Quality and Productivity Improve-ment (CQPI) at UW-Madison.

Bill was a Fellow of ASQC, as well as a Fellow of the American Statisti-cal Association (ASA) and the Ameri-can Association for the Ad-vancement of Science (AAAS). He

PAGE 2 STATISTICS DIVISION NEWSLETTER SPRING, 1987

Statistics Division Meeting at Annual Quality Congress

41st Annual Quality Congress

The Minnesota Section Host Com-mittee is personally inviting you to attend the 41st AQC. The Congress will be held May 4-6, 1987 at the Hyatt Regency Hotel in downtown Minneap-olis. Registration costs are the same as the 40th AQC. Again a fine slate of speakers and plant tours are on the program. Please refer to the December issue of Quality Progress for complete program and registration information. Additional copies are available from ASQC Headquarters. Request one for your boss, suppliers or other col-leagues by calling ASQC at (414) 272-8575. Plan now to attend the Congress and visit our beautiful city.

From the Editor The minipaper in this issue is a

reprint of a paper by Gerry Hahn. We welcome your ideas and contributions for future minipapers. Please submit papers to Dr. John Orban, Battelle Columbus Laboratories, 505 King Ave., Columbus, OH 43201-2693.

All members of the Statistics Divi-sion are invited to attend the Division's annual meeting at the 41st Annual Quality Congress in Minneapolis. The meeting will be held at 5:30 p.m. on Monday, May 4. The meeting is sched-uled to adjourn in time for members to also be able to attend the Chemical & Process Industries Division's annual dinner meeting as well.

Members are also reminded that the Division will also host a hospitality suite at the Annual Quality Congress. Times and location will be posted at the Statistics Division booth in the exhibits area, Please stop by "for a bit of cheer" and meet your colleagues.

SPRING, 1987

STATISTICS DIVISION NEWSLETTER PAGE 3

Statistics Division ea' Cosponsors

Technical Meeting In Santa Clara

ASA CAP Committee Co-Sponsors "Values" Workshop by Doug Zahn and Lynn Hare

The Statistics Division joined with the Automotive Division and the Santa Clara Valley Section's Statistical Task Group in sponsoring a talk on "the Taguchi Approach to Parameter Design", on Monday, August 11, 1986. Over 200 persons attended to hear Diane M. Byrne of Eaton Corporation give this talk, which had won the award as best technical paper at the Annual Quality Congress (AQC) in Ana-heim. Thanks go to Diane Schreiner and Karin Beumer of Rolm Corpora-tion for arranging for this seminar and the Statistics Division sponsorship. Karin and Diane are no strangers to Sta-tistics Division activities; they also helped coordinate the Statistics Divi-sion short course prior to the AQC in Anaheim.

A videotape of this talk was made and is currently being edited for gen-eral release by Diane Byrne. A prelimi-nary copy of the videotape was used by Region 5 Councilor Steve Bailey at an "Early Bird" session prior to an ASQC Delaware Section dinner meet-ing (jointly sponsored by the ASQC Statistics Division) on March 3, 1987, at which Steve spoke on "Robust Prod-uct Design".

ASQC Quality Month Activities

The ASQC Statistics Division will again this year join with the American Statistical Association's Committee on Quality and Productivity to promote joint ASQC/ASA activity during Quality Month in October, 1987. In particular, local ASA chapters are encouraged to hold a quality-related monthly meeting during October, perhaps joint with the ASQC Statistics Division.

Thirty people attended the work-shop entitled "Using Values to Improve the Effectiveness of Statisticians" which was held January 6 and 7, just prior to the ASA Winter Conference in Orlando, Florida. The workshop, co-sponsored by the ASA Q&P Committee and the ASQC Statistics Division con-sisted of four presentations.

The first was given by Mr. George Dorman, Vice President of Corporate Quality, Westinghouse Electric Corpo-ration. He spoke about "What Corpo-rate Management Expects from Statisticians" in the Total Quality envi-ronment. He listed five major expecta-tions:

1. assist in identifying processes and operations where statistical meth-ods can effectively contribute to increases in quality and produc-tivity,

2. help managers increase their sta-tistical literacy,

3. provide training in the use of sta-tistical tools,

4. help others in applying statistics, 5. provide a resource for innovation

and advanced strategic and deci-sion-making applications of statis-tics as well as a communications link for comparing statistical notes across the organization.

He also said that an effective manager does not expect statisticians to:

1. spend most of their time fighting fires,

2. be personally involved in every statistical effort across the organi-zation,

3. be an expert in every field of application for statistics,

4. install a statistics culture by them-selves,

5. accomplish miracles overnight. Dr. Blan Godfrey, Department Head, Quality Theory and Technology Department, AT&T Bell Laboratories, also Chairman of the ASA Q&P Corn-

mittee, then spoke about "Statistics at AT&T Bell Labs." He described how the success of the Bell Labs statistics effort evolved. First, he wanted com-mitment from his management, and got it but nothing else. Then he wanted their involvement, got it, and things got worse. He learned that he needed to provide education, to give management a vision for the future contrasted with the current state of affairs and then show them how to improve. Being aware of executives' goals greatly facilitates this process. Management wants to increase sales and market share, and drive down costs. This creates an opportunity for statisticians to show how improved quality can increase market share and cut costs. Of course, executives want to know how much recommended quality programs cost and how to implement them. This is the next sale to make, and Blan gave several tips:

1. Get resumes of top executives to learn how they think and what they value.

2. Preview a presentation to an exec-utive with members of his or her staff.

3. Speak to those who have given talks to the executive previously to determine what possible reac-tions to expect.

4. Read reports executives read and write.

5. Read what executives read (Wall Street Journal, Harvard Business Review, etc.).

6. Find out what previous informa-tion executives have heard in quality seminars.

Blan also gave some tips on working more effectively with executives:

1. Ask an executive to give a talk on quality - he or she will learn a lot about it!

2. Use in-house articles on quality. Continued on Page 4

PAGE 4

STATISTICS DIVISION NEWSLETTER SPRING, 1987

"Values" Workshop Continued from Page 3

3. Quote other top executives.

4. Minimize "quantiphobia." Avoid jargon, even the word "statistics"

5. Understand the fears of middle managers: computers and statis-tics can eliminate a lot of their decision making.

6. Realize the need to change statisti-cians' stereotypic image: impracti-cal, unrealistic, living in our own world, prima donna, telling oth-ers how to do their job.

In an evening session, Mr. Dave Flu-harty, Statistical Analyst, Ford Motor Company, took on a Herculean task. Dave, an MBA turned statistician, pro-vided "A Brief Review of Business Analysis Tools" in an hour and a half. Dave discussed:

1. the importance of a healthy bal-ance sheet and profit and loss statement,

2. the four P's (product, price, pro-motion and place) of marketing and the product life cycle,

3. strategic planning and cost con-trol, and

4. cash flow analysis for project selection.

In presenting his material, Dave gave a clear picture of the many demands encountered by the business manager and emphasized how important it is for statisticians to understand the man-ager's concerns. The major portion of the workshop was presented by Dr. Charles E. Dwyer, Director and Senior Research Analyst for the Management and Beha-viorial Science Center of the Wharton School of the University of Pennsyl-vania Chuck began the workshop by asserting that you can get anyone to do anything you want. He immediately added the critical qualifier, "if you are willing to pay the price." One of the major ways we achieve results is to ask others to do things. On the back of Chuck's business card is the statement he considers whenever making a request:

Never expect anyone to engage in a behavior that serves your values unless you give them adequate rea-son to do so.

He presented a model for the process that someone, call him Carl, uses (though he may not be consciously aware of it) in assessing the value to him of complying with a request from someone else, call her Rita. Carl first examines whether he con-siders himself to be capable of comply-ing with this request. If he thinks he is capable, then he considers the follow-ing steps. If he does not think he is capable, then in one way or another, he says "No" to the request. Note that what is critical here is Carl's perception of his capability rather than his actual capability. Second, Carl considers the value he may receive by complying with the request. The value Carl sees depends in part on how well Rita has helped Carl to see that he will receive results that he values by complying with her request. Third, Carl assesses the probability that he will achieve that value, so that he can compute the expected value to be gained. This probability depends in part on how much Carl trusts Rita. Fourth, Carl subtracts the costs to him of complying with the request. These are expenses such as time, money, or inconvenience that he knows he will incur if he complies. Fifth, Carl assesses the risk associated with complying with the request. The risk in complying with the request is the set of expenses which he might incur. An example here is embarrass-ment if the project is a failure. Chuck suggested that Carl then sub-tracts his cost and risk from the expected value to be achieved. If the result is positive, he says "Yes" to the request; if it s negative, he says "No." Suppose Carl's initial response is "No." If Rita is willing to pay the price in time, discomfort, ego, study, etc., there are many things she can do to shift Carl's answer from negative to positive: increase the value to him, increase his assessment of the proba-bility of receiving that value (possibly by giving him promises in writing),

decrease the cost by offering to do some of the work herself, decrease the risk by involving others so that the risk is shared.

Chuck's work blends concepts from philosophy, epistemology, cognitive and development psychology, cultural anthropology, corporate culture, and management history. An article which reports on the use of his model in mar-ital therapy is "Overcoming Ambiva-lence Through the Use of Values Analysis" by Larry Hof and Charles E. Dwyer, The American Journal of Fam-ily Therapy, Vol. 10, No. 1, 1982, pp. 17-26.

Comments from workshop partici-pants were positive. Most people regarded Chuck's comments on behav-ior, perception and values and his five-part model as being very relevant to their work. The remaining material provided needed background.

Plans for a similar workshop, per-haps at an annual ASA or ASQC meet-ing are being considered. Those interested should contact Lynne Hare.

Thanks to Joe Voelkel of Allied Cor-poration for excellent notetaking.

Chairman's Message Continued from Page 1

Fellow Nominees

The Division is actively seeking names of individuals to he nominated for Fellow of ASQC. If you would like to nominate someone, please contact our Examining Committee Chairman, Bob Perry at (314) 925-4428.

Minnesota Hospitality

On behalf of the Division, I would like to invite all Division members to stop by the Statistics Division hospital-ity suite and booth at the Annual Qual-ity Congress in Minneapolis. The suite will likely be in the Hyatt Regency and be open (at least) on Monday and Tues-day evenings. Exact times and loca-tions will be posted at the Statistics Division booth in the exhibits area. The officers and councilors would really like to meet and talk with you so that we can more fully direct Division activities toward meeting your needs.

AN

2

0 190

so%

4

0 STANDARD.OPERATING CONDITION X CYCLE 1 EVOP RUNS

200 210 220 230 240

MOLD TEMPERATURE (°F)

Figure 1. Contour lines showing effect on present yield of ram speed and mold ternperture.

SPRING, 1987

STATISTICS DIVISION NEWSLETTER

PAGE 5

Mini Paper Reprinted with permission from CHEMTECH, March, 1976 Page 204-206 (rk

PO4

Process Improvement Using Evolutionary Operation by Gerald J. Hahn

Is your process operating at opti-mum conditions?

If not, maybe Evolutionary Opera-tion (EVOP) can help you. EVOP involves a deliberate program of replacing the static operation of a plant by a scheme whereby some of the important process variables are per-turbed slightly. The effect of these per-turbations upon product performance is noted and the process is shifted to obtain product improvement. The pro-cedure is then continued to yield fur-ther gains.

This column provides a brief description of EVOP and the processes for which it is suited. A subsequent column will deal with a variation of the standard approach, known as sim-plex EVOP.

Example of use of EVOP Consider a molding process for

which the process yield—i.e., the per-cent product within specifications, is to be maximized. Laboratory experi-ments have shown that ram speed and mold temperature affect yield and have resulted in the decision to run the process at the standard condition of a ram speed of 4 ft/s and a mold temper-ature of 200 °F. This condition, how-ever, is not optimum in the plant since the scaled-up production environment differs from that in the laboratory. Assume that, unknown to the process engineer, the effect of ram speed and mold temperature on average yield, with other variables held constant, is actually as shown by the contour plot in Figure 1. For example, this diagram indicates that a long-term average yield of 90% results at each of the following conditions:

• Ram speed, 5.5 ft/s; mold tempera-/Ph' ture, 210 °F.

• Ram speed, 6.2 ft/s; mold tempera-ture, 220 °F.

• Ram speed, 9.0 ft/s; mold tempera-ture, 220 °E

The optimum average yield of 93% is obtained with a ram speed of 7.2 ft/s and a mold temperature of 213 °F. The current standard condition (ram speed 4 ft/s and mold temperature 200 °F) results in a yield of only 85%.

EVOP is to be used to learn how to improve process yield; that is, to move from the current 85% yield point in the contour plot to a more favorable condition. This is done by running repeat iterations involving slight per-turbations of ram speed and mold tem-perature around the current standard condition, as well as the standard con-dition. These iterations are continued until a statistically significant result is determined. Each such iteration involves the standard condition (ram speed 4 ft/s and mold temperature 200 °F) and the following four added conditions (see X's in Figure 1):

• Ram speed, 3.5 ft/s; mold tempera-ture, 195 °F

• Ram speed, 3.5 ft/s; mold tempera-ture, 205°F

• Ram speed, 4.5 ft/s; mold tempera-ture, 195°F

• Ram speed, 4.5 ft/s; mold tempera-ture, 205°F

As can be seen from Figure 1, an itera-tion involves the corners of a rectangle (perturbed conditions) and the point in the center of the rectangle (standard condition). The long-term yield at some of the perturbed conditions is slightly better than at the current standard condition and slightly worse at other conditions. This, of course, is not immediately evident due to chance fluctuations. However, after a sufficient number of iterations, a statistically sig-nificant difference will be established between the responses; at that time, the fact that the highest percent yield is obtained with a ram speed of 4.5 ft/s and a mold temperature of 205 °F will also be noted. The process center is then shifted from the current standard condition to this new condition (which results in a long-term yield of 87%).

Continued on Page 6

RAM SPEED (ttlsec) 12 —

NUMBERS INDICATE DIFFERENT EVOP CYCLES CIRCLES INDICATE CENTER OF EVOP CYCLE

1 190

200 210 220 230 240 MOLD TEMPERATURE (°F)

Figure 2. Contour lines plus seven EVOP cycles.

PAGE 6 STATISTICS DIVISION NEWSLETTER

SPRING, 1987

A new cycle of iterations is then con-ducted around the new process center. In addition to the center condition of ram speed of 4.5 ft/s and mold temper-ature of 205f, this involves a rectangle consisting of the following four per-turbed conditions (Figure 2):

• Ram speed, 4 ft/s; mold tempera-ture, 200f

• Ram speed, 5 ft/s; mold tempera-ture, 200f

• Ram speed, 4 ft/s; mold tempera-ture, 210,F

• Ram speed, 5 ft/s; mold tempera-ture, 210,F

This second cycle of iterations should eventually reveal a new best condition, which is established as the new proc-ess center, and the preceding sequence is repeated. This procedure is contin-ued until no further improvements appear to result. In the example described here, this happens after a total of seven EVOP cycles (Figure 2). By this time, the center condition has moved to a ram speed of 6.5 ft/s and a mold temperature of 215F, resulting in a yield of about 92.5% this is very close to the true (unknown) optimum yield of 93%. Thus, the EVOP pro-gram has succeeded in improving the process yield from 85% to 92.5%. At this point, no important improvements appear to be attainable by further per-turbations of ram speed and mold tem-perature. These variables are now held constant at the center condition and the EVOP program is continued by introducing perturbations in other process variables to attain further improvements and information.

Basic viewpoint

Evolutionary Operation was intro-duced about 20 years ago by Professor George Box, with the viewpoint that a manufacturing process, in addition to providing a product, should yield information to improve that product. The basic philosophy is that the proc-ess engineer need understand the effect on performance of changes in the process variables. This helps him improve performance immediately. It also enables him to move more rapidly to counteract the effect of fluctuations

in uncontrolled manufacturing varia-bles that might occur at a later time.

An important requirement is that EVOP be simple to use by production personnel on a routine basis. Thus the required calculations and procedures are very elementary (1).

Evolutionary operation and the design of experiments

Evolutionary Operation differs from planned statistical experimentation (2) in several ways. An EVOP program is conducted on the manufacturing floor during actual production. In contrast, a planned experiment is usually per-formed in a laboratory during product development, and often a scaled-down version of the product line is involved. A planned experiment thus can involve relatively large perturbations of the process variables to determine their effects most rapidly. The economic con-sequences of poor performance at some conditions in an experimental program are not generally of great concern.

In contrast, the perturbations intro-duced during an EVOP program are generally small; this guards against a substantial deterioration in perform-ance. The resulting effects, therefore, are also small and often can be deter-

mined statistically only after a fair number of iterations. (For a different approach, see my next column dealing with simplex EVOP) Also, laboratory experiments frequently are larger and more involved than a typical EVOP program, which is administered by production personnel and must not interfere with the normal flow of pro-duction. Finally, EVOP has been pro-posed as a continuing mode of operation, or at least as a long-term program, while most experimental programs are short-term efforts.

Evolutionary operation and quality control charts

Evolutionary Operation differs from standard quality control chart proce-dures in that control charts are designed to detect out-of-control situa-tions: in contrast, a purpose of EVOP is to prevent the occurrence of such situ-ations in the first place. Both approaches can be used profitably for many production lines.

EVOP for tracking moving processes

An important use of EVOP is for applications where the relationship

Continued on Page 7

RAM SPEED (ft/sec) 14 —

PERCENT YIELD CONTOURS FEBRUARY 80% 12

10

8

80%

1 I l I i i

210 220 230 240 250 260 MOLD TEMPERATURE (OF)

4

2

0 270 280

PERCENT YIELD CONTOURS

15 FEBRUARY

190 200

Figure 3. Example of moving response surface.

SPRING, 1987

STATISTICS DIVISION NEWSLETTER

PAGE 7

between the process variables and the process response changes with time (Figure 3) due to fluctuations in raw material properties, changes in opera-tor practice, variability in ambient con-ditions, etc. The effect of such process fluctuations whose true causes may be unknown, might be compensated for by appropriate adjustment of the proc-ess variables via EVOP—as long as the process does not move more rapidly than the EVOP scheme's capability to react.

What processes lend themselves to EVOP?

Evolutionary Operation is applicable for some processes, but not for all. The following characteristics favor the use of EVOP:

• High-volume production • The potential benefits of process

improvements are large (the process is an important one and not already operating at optimum conditions)

• Important process variables can be identified

• The identified variables can be per-turbed readily

• The process stabilizes rapidly after a process change

• The process response can be rapidly obtained and measured

over, since these returns are always uncertain, the investment is specula-tive in nature, Therefore, to break even, the monetary gains resulting from EVOP must at least balance out the costs involved in its administra-tion. Second, in perturbing a process from its current condition, there is always the chance of incurring some losses. If the process is already operat-ing at an optimum, performance can deteriorate but cannot improve. Few processes, however actually operate at optimal conditions. It might also be argued that errors or misjudgments are more likely to occur when process changes are made than when such changes are not undertaken. For exam-ple, due to a reading error, temperature may inadvertently be changed by 100Instead of by the intended 10/ A recent survey (3) indicated that "reluc-tance to perturb the manufacturing process" was the most salient reason for not using EVOP.

Sources of further information

This column provides an introduc-tion to evolutionary operation. A recent book (1), directed principally at engineers and production supervisory

personnel, provides a detailed and very readable in-depth discussion of EVOP, including a description of the technical details of implementation. Numerous articles describing success-ful EVOP applications, principally in the chemical industry have also been written; many of these are referenced in the extensive bibliography in Refer-ence 1, References 3 and 4 provide a more detailed discussion of many of the points of this column; the results of a national survey, which evaluates how extensively EVOP is being applied and how it is being used, are also given.

Most EVOP applications have been in the chemical industry. However, the concept might equally well apply to other high-volume production proc-esses. One common point appears to characterize all the successful applica-tions of EVOP. The statistician, the research scientist, and others can pro-vide important technical guidance in an EVOP program. Active management support and encouragement are also important and necessary require-ments. However, these ingredients alone are not sufficient. Most crucial to the success of an EVOP program is the active commitment, participation, and

Continued on Page 8

Some pros and cons

For those processes for which it is applicable, EVOP can lead to impor-tant product improvement. The result-ing better understanding of the process, in addition to protecting against later deterioration, may also lead to improvement of related or future processes. A further benefit is the increased awareness and sense of involvement with process perform-ance by operating personnel, which comes with the use of EVOP.

A first disadvantage of EVOP is that its implementation costs time and money in training personnel, keeping

'41.6* and analyzing simple records, making process changes, etc. These costs might be quite modest relative to the potential gains, but, like most invest- ments, they precede the returns. More-

SIC THOMAS J LIPTON INC POO SYLVAN AVE '3-105

CLF. NJ '7,7632

PAGE 8

STATISTICS DIVISION NEWSLETTER SPRING, 1987

Mini Paper Continued from Page 7

leadership of those with the direct day-to-day responsibility for the manufac-turing process.

My next column will review simplex EVOP, a dynamic alternative to the scheme described above, and will pro-vide further illustrations of the use of EVOP.

About the Author GeraldJ. Hahn, who is Manager of

the Statistics Program at GE's Research and Development Center, has, since 1955, provided consulting, develop-ment, and problem-solving services on a variety of statistical problems throughout the company. He also is an adjunct professor in Union College's Graduate Studies division. Dr. hahn holds degrees from City College of

New York, Columbia University, Union College, and Rensselaer Polytechnic Institute. He has published extensively in the statistical and engineering litera-ture, and is co-author of the book "Sta-tistical Models in Engineering He is a Fellow of the American Statistical Asso-ciation.

References

(1) Box, G.E.P., Draper, N.R., "Evolutionary Opera-tion", Wiley & Sons, New York, N.Y., 1969.

(2) Mahn, G..", "Basic Considerations in Designing an Experiment", Chem. Technol„ 5, 496-8, 561-2 (1975),

(3) Hahn, G. J., Dershowitz, A. F., "Evolutionary Oper-ation—A Tutorial Review and a Critical Evaluation". 16th Annual Technical Conference, Chemical Divi-sion, American Society fur Quality Control (1969). (Copy may be obtained from authors at GE Corporate Research and Development, Schenectady, N.Y. 12345, U.S.A.)

(4) Hahn, G. J., Dershowitz, A. F., "Evolutionary Oper-ation Today—Some Survey Results and Observations", J. R. Stabs, ASSOC., Ser. C (Applied Statistics), 23, (2), 214-18 (1974).

Author's address: Bldg. 37, Room 578, General Elec-tric Co., Corporate Research & Development, Schenectady, N.Y 12345.

Membership Report—Second Half of 1986

As of the end of 1986, membership in the ASQC Statistics Division stood at 7937 versus 7919 as of June, 1986. This apparent negligible increase is actually due to two counterbalancing effects. The Division obtained 1334 NEW members (a growth of 16.8%) during the second half of 1986, while there are still about that many current members who have not renewed their membership as of yet.

Finally, a quick update, now that the January, 1987 totals have arrived. With 178 additional new members, the Sta-tistics Division has now passed the 8100 mark in membership. This means that one out of six ASQC members also belong to the Statistics Division. We feel that you get a lot of benefits for the incremental $3.50 extra per year it costs to be a new or continuing mem-ber of our division.

The ASQC Statistics Division Newslet-ter is a publication of the Statistics Divi-sion of the American Society for Quality Control.

MI Communications regarding this publication, excluding change of address, should be addressed to:

Anthony Salvia, Editor ASQC Statistics Division Newsletter Industrial Engineering Department The Pennsylvania State University Station Road Erie, PA 16563 (814)898-6338 Other communications relating to the

Statistics Division of ASQC should be addredded to:

Edward F. Mykytka, Chairman ASQC Statistics Division Industrial Engineering Department 207 Dunstan Hall Auburn University, Al. 36849 (205) 826-4340 All communications regarding mem-

bership including change of address, should be addressed to:

American Society for Quality Control 230 West Wells Street Milwaukee, WI 53203 (414)272-8575

STATISTICS DIVISION AMERICAN SOCIETY FOR QUALITY CONTROL Penn State-Behrend Station Road Erie, PA 16563

Non Profit Organization U.S. Postage

PAID Permit No. 282

Erie, Pennsylvania