|
Items in cart [1] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 69
8 STATISTICS AND OPERATIONS
RESEARCH IN THE 1990s
The discussion that follows concerns trends and challenges that departments of
statistics and operations research may face in the coming decade.
STATISTICS AND OPERATIONS RESEARCH OVERVIEW
Jayaram Sethuraman (organizers, Florida State University
IMPROVING THE COMPETITIVE POSITION OF U.S. INDUSTRY:
CHALLENGES FOR STATISTICS AND OPERATIONS RESEARCH
Jeffrey H. Hooper, AT&T Bell Laboratories
STATISTICS AND OPERATIONS RESEARCH IN THE 1990S
Gerald J. Lieberman, Stanford University
ISSUES IN STATISTICS AND OPERATIONS RESEARCH
Jerome Sacks, University of Illinois
69
OCR for page 70
OCR for page 71
STATISTICS AND OPERATIONS RESEARCH IN TO 1990S
STATICS AND OPER~ONS RESEARCH OVERVIEW
Jayaram Sethuraman
Florida State University
At the end of each decade, we consider the state of our profession. By our profession, I mean the teaching
of statistics, statistics curricula, and statistics departments. The old controversy on too much theory or too many
applications in our courses is essentially over. We now recognize that we need both the theory and the
applications.
At the present time, I see three main challenges in our curricula that need to addressed. The challenges come
from industry, computers, and the need for cross-disciplinary work. Industry has been demanding that we
produce more statisticians to help improve the quality of manufactured products and meet the challenge they
are facing from other nations. This has led us to revise our curricula and introduce new types of courses. These
courses include the design of experiments and other topics such as optimal stochastic control.
The second challenge has come from the availability of powerful and inexpensive computers to do our
calculations. Resampling techniques, bootstrapping, and many other methods have become commonplace
because of the availability of computers. Also, computers spawn new theory, and the theory leads to more
computation.
The third challenge is coming from sister departments in our universities. There has been a flight of service
courses from statistics departments to other academic departments. Additionally, some of the new develop-
ments in research in statistics are coming from these departments. Statistics departments should cooperate in
a team effort with these departments and develop appropriate curricula. There is a need for interdisciplinary
programs in statistics.
71
OCR for page 72
OCR for page 73
STATISTICS AND OPERATIONS RESEARCH IN THE 1990S
IMPROVING THE COMPETITIVE POSITION OF U.S. INDUSTRY:
CHALLENGES FOR ST~STICS AND OPERATIONS RESEARCH
leered lI. Hooper
AT&TBell Laboratories
I speak to you today in three different roles: first as a worker in a U.S. industry that is struggling to compete
in the world market; second, as a quality practitioner trying to improve our ability to compete, and finally, as
one who believes that our universities are a central part of the long-term solution to the problem of competition
that is facing U.S. industry.
My message is short and simple. The problems facing U.S. industry are very serious and threaten our ability
to compete. We have been trying to address these problems for approximately a decade. Some progress has been
made, but our progress is much too slow. We need more help and support from our universities if we are to
improve more rapidly.
Many factors affect the ability of U.S. industry to compete. However, the essential problem is simply that
there are many world-class competitors who bring products and services to market more quickly, at lower cost,
and with higher performance and reliability. Industries facing severe competitive pressure include consumer
electronics, semiconductors, automobiles, and telecommunications equipment.
As serious as these problems are today, they will be even more difficult in the future. One reason is that there
will be three powerful economic centers in the 1990s: the Pacific Rim, the European Economic Community,
and North America. The United States will no longer be in a dominant position. This will have very serious
impact, not just on our companies but on our universities as well. The reduced market share and profit margins
of many U.S. industries will result in a loss of jobs and the erosion of our standard of living.
Just as there are many causes of our problems, there are many parts to our response. However, the core of
the response adopted by many U.S. companies is total quality management. This involves a complete change
in how and what we manage in our businesses. As you can imagine, this is a very difficult thing to do.
But the idea is incredibly simple. Its essence is that we spend too much time trying to manage results and
symptoms when we should be managing what we do and how we do it. That is, we should focus on the customer
and understand his needs and expectations, align our value-adding processes to meet these customer needs and
expectations, and continuously improve the effectiveness and efficiency of all the processes by which we add
value for the customer. One result will be increased customer satisfaction. Hence, revenue and market share will
increase. Another will be decreased cost through increased efficiency. Together, these will have enormous
impact on profit.
Everyone in the company must be involved. Issues such as teamwork and enabling and empowering
individuals must be addressed. The objective is to get everyone to improve the efficiency and effectiveness of
their work processes. To accomplish this, data must be collected and analyzed on all work processes to
determine how to improve them. Everyone must understand how to do this. This means understanding the
simple concepts of planning experiments, collecting data, and analyzing data to provide fact-based rather than
opinion-based management. It means understanding the concepts of variability because all operations require
the separation of signals from noise. And it means understanding process flow charting, process analysis, and
process control and capability.
To date, entry level workers have not had the kind of knowledge and skills required. This has resulted in
73
OCR for page 74
CHAD~G THE MATHEMATICAL SCIENCES DEPARTMENT OF TO 1990S
massive training and education programs in U.S. industries. For example, at AT&T we are faced with training
280,000 people. That is far beyond the scope of any previous training and education program. Now you, in a
sense, are our supplier. We need to form a partnership to reduce the scrap and rework in the existing education
process. We do not want to continue such heavy involvement in the education business. There are many things
that must be done to form this partnership. One of the first is to understand where the problems are, what the
problems are, and the environment in which these problems have to solved. Universities need to understand
industrial needs. There is no better way to gain this understanding than through consulting relationships
involving both professors and students.
With respect to general education and training, everyone needs to be quantitatively literate. The degree of
quantitative literacy of many people coming from fields outside of mathematics and engineering is extremely
low. Many people go into symbol shock if they see "x-bar." There is a great deal of work to do here. Industry
views many of the people in this room as having the responsibility to help solve this problem.
Statistics and operations research departments should reexamine the role of service courses taught to other
departments. We need more cross-disciplinary teaching. Those not in your field need to understand how to use
your methods to solve their problems. They do not want to become specialists in your field, but they need to
understand what you have to offer and how it can help them. Ideally, your methods should be integrated into
their courses.
Universities can also help with the rework process for people who are in the workplace today. Short courses
for industry are very important. A growing number of universities are making real contributions in this way.
This is a lucrative activity. It builds trust and communication and helps the university faculty understand
industrial needs.
The last topic I wish to address is research. There are many research opportunities with industry. I am much
less pessimistic about this than Dr. Lieberman. But you will not tee successful at developing these opportunities
if you view industry as suppliers of money that you can spend as you alone see fit. Industry is looking for a
partnership where universities take the time to understand and solve long-term industrial problems. For
example, the Semiconductor Research Corporation has been a successful partnership between industry and
academe. Here the semiconductor industry is channeling over $20 million a year into focused university
research projects. This has resulted in useful long-term research programs, increased faculty understanding of
industrial problems, and students very well prepared for industrial positions. The support for this work is
growing.
Many similar opportunities exist in fields such as statistics and operations research. Industry is convinced
that quality and productivity are critical to their success. It remains for you to convince industry that your work
can improve quality and productivity. Take the view that we are a customer or a potential customer. Take the
time to understand our problems. If you can show how your work helps solve our important problems, it will
generate enthusiasm and support.
Other opportunities can be found in industry's relentless drive for new technologies. These technologies
open up new areas for your research, but without partnerships between industries and universities, you will not
know the areas of future importance. For example, intelligent networks are likely to be an immense future
business. These networks will have greatly improved capabilities for managing and moving information, but
research needs to be done in many areas of dynamic reconfiguration of these networks. Areas such as fault-
tolerant design and network reliability, nonhierarchical routing algorithms, and large distributed databases must
be addressed. The quality of the data in these databases, data on which we run our businesses, is three orders
of magnitude worse than the quality of our electronic components. Most of our databases average approximately
five percent defective, whereas our components are running about 50 defective parts per million. I know of only
a handful of people who are working on this problem. Lightwave technologies are opening new areas.
74
OCR for page 75
STATISTICS AND OPERATIONS RESEARCH IN THE 1 990S
Additionally, there are many opportunities in computer-aided engineering and computer-aided design.
In conclusion, a more effective partnership between industry and academe is essential if we are to improve
the competitive position of U.S. industry. At the same time this partnership will address the growing university
problem of obtaining long-term financial support.
75
OCR for page 76
OCR for page 77
STATISTICS AND OPERATIONS RESEARCH IN TO 1 990S
STATISTICS AND OPERA;rIONS RESEARCH IN THE 1990S
Gerald J. Lieberman
Stanford University
The mission of any department in a university is teaching and research. Departments of statistics and
operations research in the 1990s will be judged by these criteria. The 1990s will pose severe hardships on these
departments because of tight budgetary constraints, on both the academic side and the research side.
Both public and private universities are faced with tight budgets. State-supported schools are suffering
because of limited available funds earmarked for education. The needs of states for funding social programs
and the reluctance of legislatures to increase taxes result in belt tightening for education. For private universities,
the cost of academic programs continues to exceed the cost of tuition, and we are near the limits on this impor-
tant component. Therefore, senior academic administrators, such as deans and provosts, will be scrutinizing
what I call nonmainstream departments such as operations research and statistics. One does not hear of a
university without a mathematics department, but we all know of universities without statistics or operations
research departments. Therefore, what can such departments do to enhance their positions?
I have a few suggestions. There must be a genuine interest in "service course" teaching where we spread
the gospel to the "uninformed." This interest must be matched by teaching performance in these classes. There
are too many departments in every university that teach their own courses in probability, statistics, and
operations research. We cannot expect the dean to mandate that these courses not be taught in peripheral de-
partments. Indeed, in my experience, they often do a better job than mainstream departments. It is a matter of
competition. If we do the better teaching job, these departments of psychology, sociology, industrial
engineering, and electrical engineering, etc., will send their students to our elementary courses. At the same time
that we teach service courses, we must nurture our own majors, and provide them with the education we both
want. At the bachelor's and master's levels, we must train our students to go out to the real world and apply the
techniques of statistics and operations research to problems faced by government and industry. The educational
program of the students must reflect this mission. Finally, one of the most important problems of the 1990s is
the issue of affirmative action. We need more women and minorities on our faculties and in our student bodies.
To be successful in the former, we must increase the pipeline of the latter.
The research issues faced by departments in the 1 990s are intricately linked with the doctoral program. Many
of us have been fortunate enough to have research support for our graduate students. As the federal budget
tightens, and it will become even tighter than it is today, research support will be threatened. Not only does this
affect the faculty member directly, but it will place heavy pressure on graduate student support as well. As this
source of funding is diminished, the number of doctoral students will be reduced. Not only will this have an
adverse effect on the research program of the faculty member, but it will have a negative impact on the graduate
program in the department. This is one of the most difficult problems to resolve because of our long history of
federal support. Perhaps this support can be replaced by state and industrial support, but I am somewhat
· · ~
pessimistic.
I have said little about the profession in the 1 990s. Here, I am optimistic. In government and industry, Here
seems to be less need to continually justify the existence of trained statisticians and operations researchers.
Indeed they are, and will continue to be, in demand. This is the challenge for universities.
77
OCR for page 78
OCR for page 79
STATISTICS AND OPERATIONS RESEARCH IN THE 1990S
ISSUES IN STATISTICS AND OPERATIONS RESEARCH
Jerome Sacks
University of Illinois
A number of issues have been described by the previous speakers, not only in this session, but also in earlier
sessions. They have focused on education. However, they have not covered the entire panoply of educational
issues.
There is an industrial issue of making sure people in industry are trained properly. The entire defense
industry is involved in the issue of training its workers in statistics.
We are also concerned about the relevance of graduate education, relevance not only for doing research and
teaching, but also for entering industry and government. The issue of service courses for other departments and
the need for quantitative literacy throughout the university must be addressed. Additionally, the emergence of
statistics in pre-college will create a greater demand for statistics at the undergraduate level.
It would probably take twice the resources currently available to address the educational issues just
mentioned. That list does not connect with research. In research we have a shrinking federal research budget
for statistics in particular, and for mathematics in general. This has happened in pert because of policy decisions,
in part because of the federal budget, and in part because of the failure of the mathematics community to adapt
to changing realities, needs, and opportunities.
One of the realities is motion and outreach. I don't like the word "outreach." We must really "connect" to
other sciences, other departments, and industry. To connect in this arena is not a question of opportunity, it is
a question of necessity.
I do not wish to repeat what Dr. Hooper described as necessity. Just let me underline the fact about the
competitiveness of American industry. That issue alone is enough to drive all of our work.
Where does this bring the statistics community? I think it relates to the differences between the little science
attitude we have traditionally had and the big science attitude that is being pressed upon us so that we can address
existing and emerging problems. We are not conditioned to do big science. However, one cannot do little science
and expect to have an impact on nuclear fusion, on the greenhouse effect, or on anything requiring collaborative
effort with a variety of scientists and a variety of tools.
A tremendous agenda faces us in the 1 990s, and the human resources are not available. Part of the problem
is lethargy. The carrot and stick approach, mentioned yesterday by Dr. Reed, may sound drastic and unpleasant,
but we may need to employ it to drive the mathematical sciences into doing something meaningful.
There is a fundamental dilemma. We do not have enough resources to deal with the educational needs. To
divert resources into developing the human resources means neglecting the absolute necessities of research. I
do not know how that will evolve.
We do have a strategy available to us to push ahead, a strategy that is perhaps foreshadowed by the NSF'S
regional geometry institutes. We must merge the research interests with the educational and public relations
interests of our fields in a way that will reinforce or create a public image that ours is not a dull, dry, and Relevant
subject, but something that is dynamic and essential to life on this planet as we know it. If we should fail, well,
at least we can go out with a bang and not a whimper.
79
OCR for page 80
OCR for page 81
STATISTICS AND OPERATIONS RESEARCH IN THE 1990S
QUESTION-AND-ANSWER SESSION
QUESTION: There seems to have been at least one prospective model for industry/university cooperation in
biotechnology development in terms of jointly funded institutes and industry/university personnel exchange.
Do you view this as a valid model for what you would like to see in statistics?
DR. HOOPER: I have seen two models. I am somewhat familiar with the one you mention. However, I am more
familiar with two others. Semiconductor Research Corporation is a consortium of about40 U.S. semiconductor
companies, each of which contributes, according to a formula, to a fairly sizable fund in the $20 million range.
The money is apportioned to the universities to do research in problem areas important to that field. This
consortium has been extremely effective. It not only produces students who really understand current industrial
problems, but it focuses a lot of good, solid research on some of those problem needs.
A much smaller one, in more fundamental research, is concerned with superconductivity. AT&T and IBM
are entering a partnership with MIT for a major research thrust in the area of superconductivity.
I do not think there is any one model. There are many models that could work. It depends on the stage of
the technology and how fundamental the research is. There are many opportunities to use ingenuity in these
areas because the partnerships are sorely needed.
DR. SACKS: There is an emerging development in some universities: the centers sponsored by the NSF and
the DoD. Some of them have industrial components and research activity that combine a number of efforts.
Large institutes are also developing. We have one at the University of Illinois, for example, the Beckman
Institute. There is a similar center at Cal Tech. Both of them have industrial components. One sees a variety of
strategies to attempt this coalition of effort to advance in appropriate directions. Unfortunately, the mathemati-
cal sciences are not as well represented in these areas as they might be. That is due in part, perhaps, to the
pathology in our sciences and in part to the shortage of manpower.
PARTICIPANT: There is a problem with the suggestion about teaching in other areas to teach our methods. One
of the problems with operations research is that its methods are being taken over and taught by other areas.
Electrical engineers are using and teaching queuing theory. Business departments are teaching linear program-
ming. Also, statistics is being taught in psychology and sociology departments.
We face competition in working with industry from people in other disciplines who are also working with
industry. Maybe they don't know our methods as well as we do, but this is a potential problem. It is going to
impinge upon our support, our majors, and our research.
DR. HOOPER: It gets to the heart of whether you view it as bad that they think your methods are important
enough to teach themselves or whether you view them as competitors. If you view the other departments as
value-added resellers for what you have and you are providing them with additional insights into how to teach
your methods, apply your methods, or even develop new methods, then you will have more impact and not have
to do all of the teaching yourself. I think you may need to take a more business-type view of a different kind
of business relationship with them instead of viewing them as competitors. You could be suppliers to them in
a very constructive way.
81
OCR for page 82
CHAIRING TO MATHEMATICAL SCIENCES DEPARTMENT OF TO 1990S
PARTICIPANT: I was struck by Dr. Lieberman's remarks on the disparity between the agenda he has laid out
and our resources, in terms of manpower, to accomplish that agenda. That leads me to wonder if perhaps we
need a little setting of priorities as a profession. For example, there is a massive effort now to restructure the
teaching of calculus. Given the concern about basic statistics, perhaps we need to restructure basic statistics
courses. Maybe we need our own version of the David report for purposes of clarity on our part about what our
priorities are as much as for purposes of promoting a greater sense of community within the statistical sciences.
Would you like to comment on that?
DR. SACKS: You are absolutely right. The field needs that sort of thing.
82
Representative terms from entire chapter:
service courses
