Jim Cochran

In the 12th installment of the Amstat News series of interviews with ASA presidents and executive directors, we feature a discussion with Ronald L. Iman, 1994 president.

Q: Ron, thank you for taking time for this interview. You published several papers with W. J. Conover on using rank-transformed data. What prompted your early career interest in nonparametrics?

A: Jay Conover was on the faculty at Kansas State when I was in graduate school (he is now at Texas Tech). I took his nonparametrics course when his benchmark textbook Practical Nonparametric Statistics—now in the third edition—was just in manuscript form. I was taken by the subject, and I offered small suggestions to Jay during the book’s development, which eventually led to Jay becoming my major professor. The term “rank transform” originated in my dissertation at Jay’s behest. In 1980, we published a paper titled “The Rank Transform as a Bridge Between Parametric and Nonparametric Statistics” in The American Statistician. With nearly 3,000 citations to date, that paper is the second-most-cited paper in the history of TAS. It is still cited 85 times a year. A doctor told me last month that he had looked up rank transform on Google and got 61 million hits—this number is likely inflated due to search procedures, but it provides evidence of widespread use.

Q: You were employed by Sandia National Laboratories for more than two decades. What about Sandia’s work environment appealed most to you? What was the most interesting and challenging problem you worked on while at Sandia?

A: As a distinguished member of the technical staff at Sandia, I had the opportunity to work on several problems of national interest, including risk assessments for disposal of radioactive waste and nuclear power plants. I also worked on the reliability of the proposed space station. I worked with Steve Hora on the reliability of the solid rocket boosters (SRBs) on the space shuttle. This latter work was completed about three months before the Challenger space disaster. Our analyses had shown the failure rate of the SRBs to be about 2%, so we worked very hard to get failure data from NASA to compare to our analyses, but NASA was not forthcoming. I always felt that, had we been listened to, the Challenger incident might have been avoided. Incidentally, Challenger failed on the 25th flight—making the observed failure rate exactly 2%, as there are two SRBs per shuttle. I attended a congressional hearing on the Challenger incident and, even after the disaster, NASA was claiming a failure rate of only 10^-6.

The most interesting program was one that I headed to evaluate the reliability of a new soldering process that eliminated the use of ozone-depleting hydrofluorocarbons (HFCs) to clean printed circuit cards during the soldering process. This may not sound too exciting, but this is a huge business—look no further than your computer for a circuit card. There was a lot of resistance within the printed circuit card community to eliminating the use of HCFs in the cleaning process, so I formed a group of 10 organizations, several of which were military.

This program received a lot of notoriety and, in addition to receiving the EPA’s Ozone Protection Award, it became the first recipient at Sandia of the President’s Gold Award for a technical program. As an aside, Bill Clinton visited Sandia at this time during his first campaign. He was given one of our printed circuit boards, encased in epoxy, mounted on a plaque, and billed as the first “no-clean” circuit board!

The most appealing part of these programs and others was that I was given the freedom to do my own thing and I was able to make a lot of innovative contributions.

Q: In the article “Partnering for the Future of the Statistics Profession,” which you published in 1998 as part of Statistical Case Studies: A Collaboration Between Academe and Industry, you stated, “The problems we face today are too complex for any one entity to solve in isolation, and in this era of increasingly tighter budgets, academe and industry are seeking ways to leverage scarce resources. The statistics profession should flourish in such an environment!” Do you think our discipline has made good progress in this area? Do you see the recent excitement over analytics— which most define as a combination of statistics, operations research, machine learning, algorithms, and data management—as an opportunity for our discipline or a threat to our discipline?

A: The article to which you refer was an extension of my passion for partnering that I expressed in a President’s Corner article in the October 1994 issue of Amstat News titled “Partnerships Among Academe, Industry, and Government.” This article is where I first introduced the acronym SPAIG for Statistical Partnerships among Academe, Industry, and Government, which, I am pleased to say, is still going strong.

Analytics certainly seems to have become a buzzword in many constructs—for the most part, that is likely a good thing. At the same time, there could be a danger in the sense that the direct contribution of statistics may not be acknowledged or recognized when it comes to sharing the limelight.

A key point in my comment in the publication you mentioned is whether statistics as a profession is willing to recognize and adapt to different opportunities, particularly as they relate to promotion and training. To expand on this comment, we are all aware that a fundamental tenet of statistics is it can be used to optimize available resources to maximize returns—a strong and timeless selling point. As statisticians, we have all too often failed to take advantage of opportunities to make the nonstatistical community aware of what we can do. That is, we have traditionally been enamored with the science of statistics, but, all too frequently, we fail to ring the statistical bell loud and clear.

My comments in that article were intended as a clarion call to statisticians to promote the profession while also adapting our traditional training techniques, where I think great strides have been made. Let me comment on the changes I have observed in the statistics department at Kansas State, since I am a frequent visitor to the department.

Like many of us, when I was in graduate school, there was a core set of courses that did not change all that much from year to year. Statistical computing made great strides during that time, albeit in the context of Fortran and not anywhere close to software that is available today, such as Beocat. Kansas State University had joint appointments with the agriculture experiment station from the beginning of its existence, so there was always a rich interaction in the area of agriculture.

All in all, Kansas State University had a strong set of traditional statistical courses, which continue today, but the face of the department has certainly changed from those days as the coursework has been greatly expanded, as the following recent hirings indicate.

As a DVM, Nora Bello might be considered a nontraditional hire, but she has a strong statistical background, plus the credentials to not just consult, but to do collaborative research with the college of veterinary medicine. Her collaborative research has created great opportunities for her students to be directly involved in many of veterinary medicine’s research activities. It is extremely helpful that the Kansas State University administration embraces this approach to consulting/collaborative research in a way that could be described as “all in.”

One new hire offers a new dimension in social and political science applications, having come to K-State from a post-doc in political science at Princeton. Two other recent hires are coming off post-docs in public health at Yale Medical School and Michigan State, and two other recent hires have interests in large high-dimensional data applications. K-State has also expressed an interest in applications for public policies related to food production, distribution, and consumption.

Another area that will have a direct impact on K-State’s statistics department is the $1.25 billion National Bio and Agro-Defense Facility (NBAF) under construction on the Kansas State University campus. This facility is unique in the United States and mainly funded by the Department of Homeland Security. It will be used to study diseases that threaten both this country’s animal agricultural industry and public health. NBAF provides an opportunity for the statistics department to be on the ground floor of collaborative research when the facility opens, which will require some nontraditional faculty.

Q: What motivated the development of Latin hypercube sampling? Were you working on a specific problem that motivated the development of this technique?

A: My colleague Jay Conover developed Latin Hypercube Sampling (LHS) while working at Los Alamos National Laboratory in the summer of 1975 on a project to improve the efficiency of simple random sampling. Coincidentally, I began working at Sandia at that time and I had need for such an improved technique. Jay shared his paper with me, and I wrote the first software for LHS. My need was necessitated by a time-consuming computer code that simulated peak cladding temperature for a loss of coolant accident in a nuclear reactor. LHS made it possible to extract better information with far fewer computer runs. Jay’s innovation was published in Technometrics with McKay and Beckman in 1979. With nearly 5,200 citations, it has since become the most-cited article for that journal. LHS has gone on to be used worldwide in numerous applications.

In a related development, the computer codes in use had to correlate some of the input variables, so I developed a methodology for correlating such variables without destroying the marginal distributions. I was able to resolve this roadblock by using rank correlation. That technique has been widely used by commercial software developers and the article has close to 1,200 citations.

Q: Early in your career, you worked a great deal on methodologies for disposal of radioactive waste. How did you get involved in this area of research, and how did you use statistics to contribute to these efforts?

A: In the early years, I worked a lot with computer codes used for risk assessments for geologic disposal of radioactive waste, and also with nuclear power plants. These projects were sponsored by the Nuclear Regulatory Commission (NRC). The NRC requested a sensitivity analysis (SA) for one of the models without providing any working definitions or guidance. I performed what I considered to be a sensitivity analysis and traveled to the NRC to make a presentation of my results. I was not sure how the NRC would react, but they said my work was just what they wanted, and we were off and running. Not too long after that, the NRC made a similar request for an uncertainty analysis (UA). After that, we produced many papers on UA/SA applications—many of these were co-authored with my colleague Jon Helton.

Q: What differences do you see in the responsibilities you took on as ASA president in 1994 and the responsibilities Jessica Utts will take on as ASA president in 2016?

A: This is a much more complex question than it might appear and could more than likely take more than one path. I spent 10 years on the ASA Board of Directors during a 12-year period. I was on the executive board for nine of those 10 years, so I had a great opportunity to really understand all aspects of the operation of the ASA. I was on the budget committee for three years, followed by four years as ASA treasurer.

During this time, I gained an in-depth understanding of ASA finances. In my experience, an organization such as the ASA cannot move ahead into new programs without having good financial health, which was lacking. I think Jessica will find the ASA in a strong financial position today, but that was not the case years ago, when the ASA had good in-house bookkeeping capabilities but no accounting skills.

Contributing to the ASA’s financial situation was the hiring of the same auditor year after year, with the selection based on being the lowest bid and in line with what the ASA thought they could afford. This is not a criterion I would recommend when selecting an auditor. We changed auditors and, by following their recommendations and making other changes, we were able to reverse the financial situation.

In addressing the financial issues, I want to acknowledge the good support I had on the executive board from John Neter, Janet Norwood, Bob Hogg, Arnold Zellner, and Stu Hunter. John was particularly helpful in setting up funded depreciation accounts for the ASA building and equipment. These accounts later proved their worth by allowing the ASA to move to its present building without a fundraising campaign. This may sound like a lot of technical jargon, but it was absolutely critical at the time.

Q: You and your wife have established awards at Kansas State University. What motivated you to do so?

A: Rae and I started endowing statistics scholarships at Kansas State University in 1998. These scholarships have provided funding for a least one student per year since 2003. We felt this could help out the department and perhaps entice some students into the statistics profession. Based on the number of nice letters we have received over the years, it appears the scholarships have been very helpful and appreciated.

Another area we established is yearly faculty awards. Eight years ago, we started making yearly cash awards to the best teacher and the best researcher on the faculty at Kansas State University. We hoped this might be one way of honoring outstanding faculty. The honorees are selected by a committee of peers and others at KSU from a list of university-wide faculty-generated nominees. We think it is kind of neat that these awards are funded by a “statistician,” and we are pleased that they reflect well on the statistics department. The awards are endowed, so they will continue indefinitely.