Journal of the American Statistical Association

Volume 92, Number 440, December 1997

Presidential Address

We've used as a theme for this year's meeting, Shaping Statistics for Success in the 21st Century, and that is also the title of my talk tonight. Many of the sessions at this meeting are devoted to different perspectives on our future. All along my hope has been that the sum of all of these perspectives (see, e.g., Eddy et al. 1998) would help to give us a clearer view of the future and point our compasses in the right direction.

How in fact are we doing? How prepared are we for the future? As I was thinking about my own answers to these questions I was searching for a metaphor that would convey to you how I feel. I think I found it!

My metaphor is in the form of a story about a friend who awakened abruptly one night while in the middle of a deep sleep. He sat up in bed breaking out in a cold sweat. All this fuss awakened his wife who asked him "what in the world is going on?" My friend, now dripping in cold sweat, mumbled "I think I'm dead!" His wife looked at him in amazement and asked "how do you know?" He thought for a moment and answered "nothing hurts!"

Nothing hurts. Well, by that standard, I think we can say that statistics is alive and healthy!

I think you know what I mean. It's an ironical situation. The need for statistical thinking has never been greater. We're surrounded by massive data challenges that cry out for improved statistical methods and algorithms. We have a host of interdisciplinary challenges to pursue. No group is in a better position to crack open the challenges of the information age than we are. The importance of our field is reflected in the nearly 40% rise in elementary statistics course enrollments between 1990 and 1995 (Loftsgaarden and Watkins 1997). And now we have an Advanced Placement or AP exam in statistics, which is a striking acknowledgment of the central role of statistics in our lives. (Dick Schaeffer reports that around 7,000 exams were graded this year, many more than expected, and that twice that number are forecasted for 1997-98.)

At the same time, statistics seems to be struggling (see, e.g., Iman 1995). Are these the natural evolutionary struggles that will help us to evolve into a stronger discipline? Or are they dividing us unnecessarily and sapping our strength at a time when we can hardly afford it?

I don't have the answers. But I do know that many of you share my concerns that, looking ahead, our continued success is far from guaranteed. The purpose of my talk with you tonight is to discuss a few of the dimensions of these concerns and opportunities and to suggest some priorities that we can work on together.

Image Reconstruction

I put at the top of the list our image or, more precisely, image reconstruction.

As scientists, we tend to concentrate on our technical work without much regard for how others see us. Unfortunately, our image—our public statistical persona—haunts and harms us. The standard stereotypes of statistics as a dull and impenetrable subject and of a statistician as a numbers nerd are pervasive and unfair. I sometimes wonder if there is any other science that is so seriously misunderstood and narrowly pigeon-holed as statistics.

The difficulties start early when students shy away from study of statistics or put their feet in the water and are quickly turned off by what they see as statistical mumbo jumbo. As youthful perceptions mature, we are hit by a battery of prejudices that again reflect ignorance or misunderstanding of our problem solving capability.

To give you a typical example, recently ASA was invited to participate in a government-sponsored discussion of privacy issues involving databases. We were happy to do so and were ably represented by John Gardenier.

After the meeting John reflected on the other attendees and wondered why we (ASA) had been grouped with sports statistics people, Dun and Bradstreet and other commercial companies. In fact there had been a second group composed of the American Physical Society, NSF, AAAS, NOAA, etc. Apparently the organizers did not appreciate that statistics is a science! It reminded me once again of the W. H. Auden lymeric, "thou shall not sit with statisticians nor commit a social science."

What can we do to burnish our image? Well, lots of things really. But let me focus on one thought: we need to devote more energy to communicating effectively with the public. We need more of our own versions of the late Carl Sagan who did so much to popularize science. In the words of Stephen Jay Gould, (Gould 1997), "Great scientists have always produced the greatest popularizations, without compromising the integrity of subject or author." A few statistical Sagans could help us immensely!

But, just as important, we all need to pitch in to tell the story of statistics in ways that non experts can understand and appreciate what we have to offer. Communication with the public at large, in language they can understand, is a challenge for all of us. We can look to our elected leaders, our senior statesmen, the ASA office, and our Office of Scientific and Public Affairs (OSPA) Advisory Committee to set the tone, but improvement on the scale that is required will only come if we all make this a priority.

Looking ahead, image reconstruction must be one of our top priorities. It must be understood that statistics is the data science of the 21st century—essential for the proper running of government, central to decision making in industry, and a core component of modern curricula at all levels of education. I would like to see ASA make image reconstruction one of its top strategic priorities. And I suspect we may need some professional help if we are to succeed.

Outreach and Advocacy

Closely related to our need for image reconstruction is the opportunity for increased outreach and advocacy. Let me illustrate with four examples.

The first example involves our ability to respond rapidly to important issues involving statistics. Today if an issue of critical national concern involving statistics were to arrive at our door, an issue that requires immediate attention, we are not really well equipped to respond. This situation has been of concern to the ASA Board of Directors. We want to be in a position to respond thoughtfully and quickly and to involve appropriate experts in a timely fashion.

As a second example, earlier this year ASA was encouraged to establish a mechanism to provide ongoing advice and council to federal agencies on fundamental matters of economic statistics. Under the leadership of the OSPA Committee, we are exploring exactly how such a mechanism should be established.

The current debate over the role of sampling in the next census is a third example. I am proud of the proactive role we took through our blue ribbon panel as well as the contributions of other statisticians to the debate on this politically charged topic.

Fourth example. Recently we joined forces with other societies to advocate the need for strong federal support for science research. We were cosponsors of a full page ad which appeared in Roll Call, a congressional newspaper. We need to weigh such opportunities carefully but we also need to be more aggressive and proactive than perhaps we have been historically as advocates for science and advocates for statistics.

Here I am thinking of the words of Mary Woolley as expressed in a recent Science magazine editorial (Woolley 1997). She pointed out that "Lobbying work in Washington is absolutely essential for the future of science; and every scientist should be contributing through his or her professional associations, to responsible representation of the interests of science in the nation's capital."

While the focus on Washington is natural, she goes on to point out that local public outreach and advocacy activities are equally important.

Looking ahead, it is hard for me to envisage a successful future for us without extensive grass roots commitment to and involvement in outreach and advocacy. As an association and as individuals we must find effective ways to get involved. Resources must be committed. Outreach and advocacy need to be major components of our strategic plan.

Human Resources

I'd like to talk for a minute about our human resources. How are we doing in attracting talent to the field? What are we doing with that talent?

In terms of attracting students to the field, I get a very mixed picture as I talk with different statistics departments. Some appear to be doing just fine. But others are seeing a sharp drop in quality applicants to the point that one might question the viability of such departments if the trend is not reversed very soon.

One bright spot appears to be the number of women entering graduate school in statistics. Women constitute from one third to one half of the population in many cases, and the numbers are especially strong in departments which provide an applied focus such as biostatistics or industrial statistics.

As I look across the landscape of ASA, I see women playing key roles in virtually every corner of the association: as President and Vice President, as editor of JASA, as Chair of the Joint Program Committee for this meeting, as Board members, as council and committee chairs, as Chapter Presidents, and as movers and shakers in the widest sense.

Having said this, however, there are other aspects of our human resource utilization that are not nearly so gratifying to talk about. Within ASA we have a demographic profile that is rich with diversity of all kinds.

Let me just say that we can and should do a better job of reflecting this diversity in all of our activities.

One opportunity that I would like to see us grab a hold of is the idea of a senior service of statisticians. Fritz Scheuren described the opportunity in these words: "ASA needs to find a way to set up a wide area network for statisticians whose workplace associations are irregular. Without this, too much talent, knowledge, and wisdom will be lost. There is so much work to do."

Sig Amster, who is now retired after working more than 35 years at Bell Labs expanded on this idea:

There seems to be a great opportunity for ASA to advance the current push for "volunteerism" by finding matches for retirees with valuable experience and local, state and federal governments with such needs.

I believe there are many retirees, including myself, who do not wish to "pound the pavements" looking for this type of work. If ASA could be the catalyst to bring the two parties together, there should be much positive publicity for the organization and much benefit to the participants.

Looking ahead, I believe we can do a much better job of attracting talented people into the field and then fully utilizing that talent in all of our activities throughout their professional careers and even into retirement.

International Relations

An intriguing feature of our demographic profile is the fact that about 10% of our members reside outside the U.S.

In recognition of the importance of our international constituency, we have a dedicated position on the Board of Directors for an international representative. We also have a Committee on International Relations in Statistics. Its charge is to "...foster activities involving international communications, exchange, and joint development with professional statisticians, societies, and other organizations in countries throughout the world."

Our high quality publications are a magnet for members internationally. Our newest journal, the Journal of Agricultural, Biological, and Environmental Statistics, is a joint publication of ASA and the International Biometric Society. Our statistical magazine, CHANCE, a joint venture with Springer, is designed for general audiences, and we hope it will become a truly international magazine. Indeed, its editor, George Styan, lives in Canada and his editorial board includes members from Finland and New Zealand.

One of the most exciting developments at ASA during the past year has been our web site. ASA's web site can be accessed in principle as easily from Moscow or Sydney as it can from Alexandria or Anaheim. It offers a tremendous opportunity for international outreach and communication.

In fact we have some early indicators of this from our web site statistics. Let's take a look at the statistics for a roughly 100 day period starting in mid March. During this period we received over 900,000 requests for information. While the majority were naturally from the U.S., requests came from over 80 other countries around the world. The leading ones were, in order, Canada, the United Kingdom, Germany, Australia, France, and Austria; but what was even more intriguing to me was the variety of others in a very long-tailed distribution—Paraguay, Croatia, China, Norway, just to name a few of the others. Overall, more than 11% of the traffic is international, which is about the same as the percentage of international members. I speculate that this is the tip of the iceberg for us. As the amount of valuable information on our web site increases, we should see a correspondingly strong growth in interest from the international community.

I think we have an excellent foundation in place for strengthening our international ties and building closer working relationships with statistical societies world wide. But we could be doing more. Afterall, ASA is by far the largest national or international society devoted to statistics. We're three times larger than the Royal Statistical Society; three and one half times larger than The National Statistical Society of China; and nine times larger than the International Statistical Institute itself.

We can take some pride in these numbers. However, we also need to share our success with others and cultivate effective international partnerships.

Looking ahead, I believe ASA should sharpen its international focus by building upon its foundation of rock solid local and national activities and expanding its ties around the world.

Now I'd like to switch gears a bit and talk about three dimensions concerning the subject matter of statistics itself.

Holistic Statistics

The core of statistics has gradually evolved and strengthened during the 20th century. The mainstream statistics journals, such as JASA, have documented these developments very successfully (see, e.g., Billard 1997). And our statistics departments have built their programs around these core topics with very little variation.

The question I wish to raise is whether the 21st century statistics discipline should be equated so strongly to the traditional core topics and activities as they are now. Personally I prefer a more inclusive interpretation of statistics that reflects its strong interdisciplinary character.

The issue is similar to what we hear in other domains of science. It reflects the tension between narrowly focused pursuit of science versus approaches that are more holistic, emphasize synthesis, and encourage interdisciplinarity. For example, a recent article about the future of physics argues for a broadened base of research, unusually flexible tenure evaluations, and increased interdisciplinary interactions (Gruner et al. 1995).

Eric Chaisson (1997) put it this way in his description of NASA's new science vision:

To make this new program really work, scientists especially will need to become more broad-minded and less specialized—to be willing to explore (or at least tolerate some of their colleagues exploring) holistic world views systematically and synergistically, not merely undertake the reductionistic science that has been so heavily supported by funding agencies for decades and just as myopically honored by our colleges and universities. The scientific community will have to welcome synthesis as well as specialization, teaching as well as research, dissemination as well as discovery.

There is no doubt in my mind that we must work hard to preserve a strong statistics core. However, I believe the best way to do this is by surrounding it with vigorous areas of application that will nourish the core with fundamental challenges and receive in return a variety of new approaches and methodologies.

The inclusive view suggests broad-minded statisticians who are not as committed to narrow specialties. Instead, they see the world at large, including the world of statistics, holistically and do "not merely undertake reductionistic science." Rather, they are skilled at looking across boundaries and operating in fast-paced environments. They know how to solve real problems and have genuine impact.

Perhaps it is even possible to have it both ways. I think Adrian Raftery described the potential very nicely in a note he sent me earlier this year:

Statisticians can work in many areas of science, while gaining a deep knowledge of aspects of their own discipline. In a world of increasing specialization, they can combine the scope of generalists with the skills of specialists. This combination is rare and in great demand, as well as providing a lifetime of exciting and fulfilling work.

Looking ahead, I believe that an inclusive interpretation of statistics, and all that that implies for training, research, and careers is the way we should define our field for tomorrow.

New Foundations

I've been talking about a broad vision and definition of statistics. Another question is what shape the core of statistics will take and, in particular, what the fundamental underpinnings of statistics will be in the years ahead.

I believe the time has come for us to acknowledge that computer science needs to take its place alongside mathematics (and probability) as fundamental linch pins of statistics and as disciplines that undergird our research and instruction. Examples of relevant topics in computer (and computing) science include databases and database management, algorithm design, computational statistics, artificial intelligence and machine learning.

It is mainly for this reason that I invited Al Aho to be the guest speaker in this year's president's invited speaker session—to symbolize the growing importance of computer science to statistics and to encourage us to think about the need for increased interactions between our fields. I believe this is the first time that a computer scientist has filled this role. In fact, it is pretty unusual just to have a computer scientist involved in our meetings!

In addition to making selected computer science topics a part of the core training for future statisticians, I see many interdisciplinary opportunities as well. Let me highlight two opportunities that sit at the intersection of computer science and statistics.

First, I'll mention the area of software engineering. This relatively young branch of engineering has emerged as one of the critical activities of the information age. Quoting from a report of the National Research Council's (1996a) Committee on Applied and Theoretical Statistics, entitled "Statistical Software Engineering",

The development and the production of high-quality, reliable, complex computer software have become critical issues in the enormous worldwide computer technology market. The capability to efficiently engineer computer software development and production processes is central to the future economic strength, competitiveness, and national security of the United States.

The report then goes on to detail a variety of areas where statisticians and software engineers could work together productively. One of the larger challenges is statistical control of the software development process where one can envisage the sort of "demingesque" progress in software manufacturing that has been achieved in more traditional hardware manufacturing.

The second area is the current hot topic of data mining, which statisticians might reasonably think of as data analysis of very large databases. In fact, if you dig down and look at what is involved in data mining, you will find a variety of statistical components, such as statistical graphics and cluster analysis. Moreover, there is a great opportunity to bring a variety of statistical concepts to bear—modeling, sampling, robust estimation, outlier detection, dimensionality reduction, etc. Nevertheless, there are new opportunities as well, and we would be wise to pay very close attention and to become seriously involved with these developments.

In fact, if we don't, there is risk that we will be blown away by the momentum that is flowing in this direction. As Jerry Friedman pointed out in his keynote address at Interface '97, we are no longer the only game in town. Many other data oriented sciences are competing with us for customers and students.

Also it should not go unnoticed -speaking of image reconstruction!- that the very term, data mining, has captured the fancy of many people, especially in the business community. It is grabbing headlines that statisticians would kill for. The image of data mining is that something powerful is going on there. The reality? Well that may be rather different. It may even turn out that the phrase of the day in the 21st century is

"lies, damned lies, and data mining."

But for now I believe we should take advantage of the momentum before it fades into another missed opportunity for statistics.

Looking ahead, the 21st century curriculum for graduate statistics training needs to be centered around aspects of computer science as much as it is today around mathematics. The needs of the computer science and statistics communities are now lightly intersected. Let's strengthen them in ways that advantage both communities.

Data Trends

The discussion of data mining leads naturally into my next topic, trends in statistical data problems.

Trends in the types of data challenges we are facing now are excellent leading indicators of the shape that statistics will take in the early years of the 21st century. Let me just focus for the moment on the size of these problems.

Somewhat surprisingly, one trend seems to be towards smaller problems. In industry, competitive pressures are so strong, time horizons so short, and data often so expensive to collect, that we are forced into uncomfortable situations where plans and decisions need to be made based on less and less data.

But the real pressure, of course, is towards increasingly large problems and statisticians, being well schooled in variability, have a tendency to encourage things in this direction.

Moreover, in many areas of application we are seeing really unbelievable amounts of data, e.g.,

• gigabytes of telephone network traffic
• terabytes of data on global climate change.

But sheer size is not the real issue. It is the combination of size and complexity that is the challenge. These so called massive data problems (National Research Council 1996b) are often not amendable to standard statistics solutions.

While statisticians have only begun to grapple with the challenges of massive data sets, some potentially valuable approaches are already emerging such as;

• adaptive sampling (learning as you go)
• guided visualization (pan and zoom)
• reliance on approximations (forget optimality)
• distributed work (people, machines)
• divide and conquer (consolidate later)
• exploit the context (the "zeroth problem")

Colin Mallows' Fisher Lecture is all about the "zeroth problem" (Mallows 1997). References related to Peter Huber's special lecture on statistics and massive data sets are also worth singling out. See Huber 1997a and b.

Looking ahead I see challenges at both ends of the data size spectrum. However, the high-end massive ones, to paraphrase John Tucker of the National Research Council, offer the grand challenge to statistics for the 21st century.

Infrastructure

The last dimension that I want to talk about is our infrastructure.

Like any serious activity, we in statistics have our own vital needs for a strong infrastructure. I'd like to comment briefly on three of these elements.

First, NISS - the National Institute of Statistical Sciences. I find it a bit embarrassing, frankly speaking, that we have yet to pull together as a community behind NISS. Afterall, this organization was created by the national professional statistical community, including ASA, to "promote, develop and perform cross-disciplinary research in problems of national and international importance." If NISS does not ultimately flourish in one form or another, it will be an ugly black mark on our profession and a major setback for an element of our infrastructure that is very much needed for our future.

Another element of our infrastructure that is worrisome is the health of some of our academic statistics departments.

The department of statistics at the University of Rochester shut its doors on June 30. Sixty miles to the west, the statistics department at SUNY Buffalo is in a real crisis. Its heading towards a merger with another department and a dramatically reduced role in the process.

At the same time, 100 miles to the southeast of Buffalo, several separate statistics groups at Cornell University are being consolidated into a new department of statistics. And earlier this year, after a long gestation period, a new department was formed at the University of Pittsburgh.

Many other long-lived departments are so well cemented into the fabric of the university that their demise would be literally unthinkable. Perhaps we can learn from them the paths to success. You can read more about this in the July-August issue of Amstat News under the heading of the birth, life, and death of statistics departments.

To round out the discussion of infrastructure, let me turn to the central ASA office. The office is located in a four-story building on Duke Street in Alexandria, VA close to Washington, DC. It is our administrative hub for everything from publications to dues collection.

It is no secret that we have had to struggle with a range of serious problems in the ASA office over the last few years and are still a long ways from a state of office optimality. The problems have included fiscal shortfalls, turnover of key personnel, and computer software difficulties. While we have been addressing these back office problems, we are also placing increased emphasis on member services and customer care. During 1997, we initiated a quality management mode of operation which reflects the best of the spirit of W. Edwards Deming. I hope that the ASA office can now proceed quietly along its own path towards excellence of operation and that you will notice the difference.

I'd like to thank Nick Fisher, our international representative on the ASA Board, and David Marker and David Morganstein of Westat for helping to facilitate our quality management efforts.

Looking ahead, our vision for the 21st century must start with an efficient and effective infrastructure across the board. Regarding ASA I'd hope for the day when the ASA office is recognized as the standard of excellence among its peers. A tight ship. Fiscally strong. Responsive and innovative. Caring of its customers. A focal point of statistical outreach to the world at large. Likewise we must strengthen the other elements of our infrastructure because we of necessity depend on them for support and progress.

Conclusion

In this talk, at this meeting, and throughout 1997, I have been trying to focus our attention on what we need to do to position our association and profession for success in the years ahead.

This is not an idle exercise. Our position today is not terrific. We might heed the words of Satchel Paige, that long-time wily and wise pitcher:

Don't look back. Something might be gaining on you.

For us that something is at our doorstep. But we're not yet in agreement that we have a problem or what the problem really is. While we ponder and debate, opportunities—sadly—continue to pass us by.

We need some closure on these discussions. Even if we're partly wrong, which we surely will be, let's get our act together and make a plan for the future. Another important philosopher, Yogi Berra, counseled:

You've got to be careful if you don't know where you are going, because you might not get there.

Our vision needs to be multidimensional. Our image requires extensive reconstruction. We absolutely must increase our outreach and advocacy activities. We ought to better utilize our own human resources. We should bolster our international ties and increase our impact there. We need to take on an expanded, inclusive interpretation of statistics that better reflects the holistic nature of the field and the wealth of ways it can contribute to the well being of society. We need to adjust the fundamental underpinnings of statistics to include a strong role for computer science as part of the foundation of our field. We ought to deal head on with the challenges of massive data sets. (I believe this will force us to rethink much of our current statistical practice.) And, finally, the critical supporting infrastructure of statistics needs to be shaped for the 21st century. Without it we won't make it.

I hope when you leave Anaheim you will have a better sense of exactly how you can and will help. Discussions are continuing. Other dimensions are being explored. Let us keep talking and moving ahead. As the great inventor, Charles F. Kettering, explained it:

We should all be concerned about the future because we have to spend the rest of our lives there.

References

Billard, L. (1997), "A Voyage of Discovery," Journal of the American Statistical Association, 92, 1-12.

Chaisson, E. J. (1997), "NASA's New Science Vision," Science, 275, 735.

Eddy, W. F., Johnstone, I. M., Kettenring, J. R., Keller-McNulty, S., Kirkendall, N. J., McDonald, G. C., and Moore, D. S. (1998), "Shaping Statistics for Success in the 21st Century: A Panel Discussion," Technical Report I-98-1, Kansas State University, Department of Statistics.

Gould, S. J. (1997), "Bright Star among Billions,"Science, 275, 599.

Gruner, S. M., Langer, J. S., Nelson, P., and Vogel, V. (1995), "What Future Will We Choose for Physics?," Physics Today, 25-30.

Huber, P. J. (1997a), "Strategy Issues in Data Analysis," in Proceedings of the Conference on Statistical Science Honoring the Bicentennial of Stefano Franscini's Birth, (eds. C. Malaguerra, S. Morgenthaler, and E. Ronchetti), Basel: Birkhauser Verlag, to appear.

Huber, P. J. (1997b), "From Large to Huge: A Statistician's Reactions to KDD & DM," Unpublished paper.

Iman, R. L. (1995), "New Paradigms for the Statistical Profession," Journal of the American Statistical Association, 90, 1-6.

Loftsgaarden, D.O., and Watkins, A. E. (1997), "Statistics Teaching in Colleges and Universities: Courses, Instructors, and Degrees in Fall 1995," The American Statistician, to appear.

Mallows, C. L. (1997), "The Zeroth Problem," Unpublished paper.

National Research Council (1996a), Statistical Software Engineering, Washington, DC: National Academy Press.

National Research Council (1996b), Massive Data Sets, National Academy Press: Washington, DC.

Woolley, M. (1997), "The Comfort Zone," Science, 275, 1243.

Modified Nov. 19, 1997