An Interview with Janet Wittes, President of Statistics Collaborative
CHANCE invited Janet Wittes, founder and president of Statistics Collaborative, Inc., to talk with Executive Editor Scott Evans. Wittes is a Fellow of the ASA, Society for Clinical Trials (SCT), and American Association for the Advancement of Science. She is also an elected member of the International Statistical Institute. She is a past president of the SCT and Eastern North American Region of the International Biometric Society. In 2006, she received the Janet L. Norwood Award for outstanding achievement by a woman in the statistical sciences. She earned her AB in mathematics from Radcliffe College (1964) and her MA and PhD in statistics from Harvard University (1965, 1970).
Evans talks with Wittes about the challenges of being a woman pursuing an education and developing a career in statistics in the 1960s, clinical trials, the Women’s Health Initiative, and the future of women in statistics.
Some people hate the very name of statistics, but I find them full of beauty and interest.
Whenever they are not brutalized, but delicately handled by the higher methods, and are warily interpreted, their power of dealing with complicated phenomena is extraordinary.
They are the only tools by which an opening can be cut through the formidable thicket of difficulties that bars the path of those who pursue the Science of Man … [and Women].
~ Francis Galton, “The Master Builder of the Modern Theory of Statistics”
(Editor’s Note: The last two words are attributed to Janet Wittes and Scott Evans.)
Evans: Why did you become a statistician?
Wittes: I entered statistics through the back door, which I suspect is the route by which most people enter statistics. My father was a chemist and I had intended to follow in his footsteps. I was an undergraduate at Radcliffe, which was then the women’s school at Harvard.
Concentrators (or “majors” in other schools) in chemistry and biochemistry were required to take nearly the same courses, but those who concentrated in biochemistry received a valuable benefit: they were assigned a tutor in their sophomore year with whom they would meet weekly. To take advantage of this unique educational experience, I chose to concentrate in biochemistry. My tutor was Professor John Edsall, whose specialty was protein biochemistry.
At the tutorial sessions, Edsall would ask me questions to gauge my interest in the field and my understanding of it. Each week, he would give me some assigned readings that we would discuss at next meeting. The first readings addressed general topics in biology or chemistry. Edsall quickly realized that I was not interested in what experiments showed, but how one made inferences from experiments. He then gave me a book describing J. Willard Gibb’s phase rule, a mathematical formulation that allows one to predict the number of stable states of matter, or phases that may exist in equilibrium for a particular system.
I was especially fascinated by the “triple point”—the point in the temperature-pressure gradient where matter could be solid, liquid, and gas. The revelation that H20 had a mathematically described triple point—where ice, liquid, and vapor coexisted simultaneously—was breathtaking to me. Relieved that he had finally found something that appeared to excite me, Edsall said something like, “I will give you another book to read. See what you think.” That book was Henderson’s The Fitness of the Environment.
I found that book revelatory as well. While the Phase Rule described the nature of matter itself, Henderson described the relationship of chemical substance to life. Its message was tied to the Phase Rule, for it emphasized the unique quality of H20, where the solid phase (ice) was less dense than the liquid phase (water), rendering oceans fit for living beings.
After several months of meetings, Edsall said to me, “You’re not really cut out to be an experimental scientist. Try this.” He assigned me Moroney’s Facts from Figures. After reading just a few introductory chapters—the ones that described, among other issues, bias, fallacious representation of data, the fact that the arithmetic average could hide a host of properties of distributions, clusters—what I found especially refreshing was his emphasis on uncertainty, an emphasis that seemed to me to be a true representation of many aspects of life. I realized that my tutor had introduced me to what I wanted to do. At our next meeting I announced, “I want to be a statistician.”
I switched into math with the view of going into statistics and never looked back. But it was very important to have had a background in experimental science because understanding the process of experimentation has been invaluable in thinking about how to design clinical trials.
I was truly lucky to have been paired with Edsall; had he not been so astute about my interests, I may never have landed in statistics. I regret that I never thanked him for sending me on a path that I have loved.
Evans: How did you choose to go into biostatistics specifically?
Wittes: When I switched into statistics, I was interested in many fields that would benefit from statistical thinking, especially experimental science (as long as I didn’t have to be in the lab), anthropology, and law.
On entering graduate school, I needed fellowship support. Most applications for fellowships were very long, but this was the spring of my senior year and I had to study for finals. I didn’t have time to write long essays about why I wanted to study statistics. So I looked for the shortest application. The Public Health Service (PHS) required the applicant to fill out only a single page for its traineeship. The clincher was the applicant had to attest to interest in fields relevant to the PHS. So I decided to learn about public health.
The traineeship was generous; it supported me for my entire graduate career and, because I was the breadwinner of the family (the “bread” being the traineeship itself), it even paid a stipend for my husband, who was then at Harvard Medical School, and later the children.
In the interest of being responsive to the requirements of the fellowship, I enrolled in several courses at the Harvard School of Public Health (e.g., epidemiology, tropical public health, cultural aspects of disease) and was fascinated by them. Again, once I made the choice, I never looked back.
Evans: In this issue of CHANCE, we’re celebrating women and statistics. Recently, you and I were at a meeting advising a large pharmaceutical company regarding an upcoming FDA Advisory Committee meeting. It was a fairly good-sized room and many people were there. At the front of the room, there was the “adult table” consisting of perhaps 15 people, primarily executives and senior-level people associated with the company, and the advisors (you and I among them). We noticed that you were the only woman at the table and I recall that you made a joke about that when we were introducing ourselves.
Wittes: At that table, one by one, each person introduced himself and proudly described his many achievements. When it was my turn to introduce myself, I just said, “I’m the token woman.” The group burst out in nervous laughter. The people at the company were very embarrassed and one of them said, “No, no, you’re not the token woman.” It’s taken a long time to be able to make those kinds of jokes and feel really comfortable.
Evans: In the past, you have told me a few stories about the challenges of being a woman pursuing an education and developing a career in statistics. Perhaps you could share what it was like being a statistics student in the 1960s when there wasn’t such a level playing field for women.
Wittes: My background differed from many other women at that time because many of the women in my family had advanced degrees. My father didn’t understand why intelligent women stopped their education at college. Two of his aunts were dentists. In fact, his paternal grandmother had gone to medical school in Poland. But because she was Jewish, she couldn’t get a medical degree. She was a licensed midwife, but was basically the physician for the area. So my father encouraged my mother to get a PhD in psychology. Thus, while schools weren’t very used to women getting advanced degrees, my family always expected that I would go to graduate school.
When we were seniors in college, Bob and I decided to get married as soon as we graduated. It never occurred to me that we should apply to schools at the same time. Instead, he would apply to medical school and I would apply to graduate school in the city where he would decide to go to school.
He chose to go to Harvard Medical School, and the only school in the area that had a statistics department at the time was Harvard.
From the very beginning, I had problems in my department. The traineeship application required me to tell the PHS by April 1 that the school had admitted me. The official date of Harvard’s acceptance was April 15. So in March, I went to the department chairman and said, “You don’t have to tell me whether you admitted me or not. But if you would just fill this form out by April 1 if you accepted me or rip it up if you didn’t, then I will be eligible for the traineeship.” The answer: “No. That’s not our policy.” I later learned that the department did fill out similar forms early for two of the men who were in my class.
So I decided to call the United States government. Remember, there was no Internet then; to find the right person, you just had to call many different offices. I managed to speak to a real person and I told the story. He laughed and said, “If you get in, we’ll give you the traineeship.” It was as simple as that. I was impressed with how nonbureaucratic the government was and how inflexible Harvard was. It was that experience that made me a fan of the federal government.
Evans: I have heard some of your fellow graduate students teasing you about tea. Is that a story, too?
Wittes: Oh that! The department held a seminar every Tuesday at 4 p.m. Our graduate school class had 10 students; I was the only woman. My job, assigned by the department, was to purchase the cookies (the department did pay for them), arrange them on platters, and brew the coffee and tea. When the refreshment period was over, everybody else would leave to go to the seminar, but I was instructed to stay behind to wash the dishes. So, each week, I would walk into the seminar 10 to 15 minutes late having no clue about the thrust of the topic. At the end of the session, the chairman of the department would ask me a question about the seminar, and I could never answer him because I had missed the crucial introductory paragraphs. Finally, Steve Fosburg, one of my fellow students, blurted out, “Why do you do this to her? You make her do the dishes so she has to come in late. Then you grill her about the seminar.” I just loved him for that.
Evans: Any other stories about being a woman graduate student?
Wittes: Here’s the worst one. I have a friend, Laura Eisen, who was a graduate student in chemistry when I was a graduate student. She and I both had fellowships that required us to register for summer school even though we didn’t have to take courses because we were just writing our theses. In July 1968, we went together to sign up for summer school. My daughter was then six weeks old, Laura’s son was 13 months old, and she was eight months pregnant. Her Dalmatian, Wink, came with us.
Imagine the scene: Laura, Wink, and I are standing in line in the hot sun outside of Memorial Hall in Cambridge. Best estimate of time to reach the front of the line: an hour and a half. We are both holding crying babies. Pregnant Laura is feeling a bit dizzy. Wink is panting, obviously thirsty.
I say to Laura, “Let me go up front and tell the people at the desk our predicament. I’m sure they’ll let us cut in line.”
I go up to the desk, which is manned by students. I’m holding my hungry, screaming infant. I report that my friend is in the hot sun with her baby and she is pregnant. The answer: “You two stand in line like everybody else.” I return to Laura in shock.
We decide to go directly to the registrar’s office, expecting more understanding there. The upshot: That office also insists that we stand in line like everyone else.
I call the government again and tell it my story. The sympathetic nonbureaucratic person on the phone says, “Yes, the PHS will continue to pay your traineeship, even if you are not registered for the summer.”
I found the whole episode incredible. You could say that Harvard was treating women no differently from the way they would have treated pregnant men. But men don’t get pregnant and they don’t come to lines in the hot sun with babies and dogs.
On the other hand, many men were terrific to me; the men in my family were totally supportive. My paternal grandfather, my father, my husband, and my brothers all encouraged me. The nine male members of my graduate class were true colleagues, and I am friendly with several of them even today.
I was friendly with the men in classes ahead of me, especially Jonas Ellenberg, Barry Margolin, and Samprit Chatterjee. My thesis advisor, Ted Colton, was consistently helpful. I would bring the baby to his office when we would talk about the progress of my dissertation. He had a professional wife and young children at the time, so he understood the issues women faced. Later, when Ted appointed me to ENAR’s Regional Advisory Committee, Judy O’Fallon, who was then a member said, “It’s so good that you’re appointing a woman.” Ted’s response was, “Oh, I didn’t realize Janet was a woman.” That was a great response. He just treated me like a person.
So it wasn’t all bad. In fact, a lot of being a woman in statistics was really very good, even as a graduate student. But we women had to fight battles in those days. Sometimes I didn’t even recognize what was happening until I realized my experiences in school weren’t happening to the males. The department of biostatistics at the school of public health, where I spent a lot of my time, was very encouraging. I don’t know whether it was the particular people there, the time, or that public health always had many more women. While my class at the department of statistics had only one women in 10 and no female professor among the five or six in the department, the biostatistics classes even then had a much higher percentage of women and a very much higher percentage of women faculty (two of four). In fact, there were few women in hard science or math or statistics in those years. I found a very different feeling on the two sides of the river. Note: For those of you unfamiliar with Harvard geography, the statistics department is in Cambridge, the biostatistics department is in Boston, and the Charles River separates the two cities.
Evans: Do you have female colleagues?
Wittes: Yes, many, and they have been important to me. Perhaps the two most important over the years have been Judy Goldberg, whom I met when we were both in graduate school on opposite sides of the river, and Susan Ellenberg, whom I met when she was an undergraduate. She and I worked together with Jerry Cornfield.
Evans: You have described your experience as a woman in graduate school and in the profession. Are there aspects of your work that relate to women’s health? In particular, can you talk about the role you played in the Women’s Health Initiative?
Wittes: In general, I do not select studies to work on just because they do or do not deal with women’s health. The Women’s Health Initiative (WHI) was a life-changing event for me. But at the recommendation of Bill Harlan, another man who was influential to me, I became chair of the Data Safety Monitoring Board for that study. I, just as I suspect all members of the committee, started our involvement believing that taking estrogen after menopause would reduce the probability of contracting heart disease and would make one generally healthier. Of course, as you know, the results of the study were quite different from what had been hypothesized. The studies reported many results, the most surprising of which was that exogenous post-menopausal hormones did not provide benefit to the cardiovascular system. To have been involved in a study that shows something really important, and that contradicts everything most people believed, teaches many lessons about the nature of evidence and belief.
Later, I heard from people who said they had always known that post-menopausal estrogen was likely to be on balance harmful, rather than beneficial. We were trying to make a recommendation on the basis of data, as the data were emerging, that would likely change the behavior of women in general. The experience was exciting and sobering.
The board had 10 members representing many disciplines. The study investigated not only postmenopausal hormone therapy, but also diet intervention and calcium. Therefore the committee membership included expertise in diet, coronary disease, strokes, bone, nutrition, psychology, breast cancer, and ethics. We had to explain to each other what we understood from our individual professional backgrounds and expertise and how that impacted our way of looking at the data. We gained a tremendous amount of respect for each other over the years.
We learned to think in each other’s heads. We gained a tremendous amount from other people, from looking at data and asking ourselves what the data were trying to tell us. We struggled with how and when to report the data to the public. If we announced results too early, people would be unlikely to believe the data. But reporting too late would lead to harming people.
Working that trial was very challenging but fascinating. The experience confirmed for me the need for randomized trials with clinical outcomes. Reliance on the epidemiologic data for clinical outcomes or randomized data with surrogate outcomes could lead to incorrect inference.
Evans: What did you do after graduate school?
Wittes: Bob completed his internship in 1970, the year I earned my PhD. This was during the Vietnam War. All physicians had to serve the United States in one way or another. Bob went to the National Cancer Institute (NCI) as a so-called “Yellow Beret” member of the Commissioned Corps of the Public Health Service. I was fortunate to find a half-time job with Jerry Cornfield dealing with the properties of what he called “relative betting odds.”
Jerry was a wonderful mentor. Sometimes I would come with the babies (by then there were two) and he would bring his granddaughter. I’m sure you know how inventive and brilliant he was; you may not know how very warm he was personally.
After Bob finished his two years at the NCI, we moved to New York. He went to Memorial Hospital and I was part time at Columbia with the department of epidemiology. After about two years there, I asked for half-time benefits, but Columbia gave no benefits to part-timers. That’s when I decided to find a full-time job, but one that did not require my being away from the kids.
There was an opening at Hunter in the math department for a statistician. I knew that job would allow me a full-time position with benefits, and yet because I was teaching, I could be home much of the day. I was very fortunate to be offered that position. I stayed in the math department for seven years. I loved working with the students and I learned a huge amount of statistics by teaching.
While I was at Hunter, I had a third baby. I used to take him to my office and the students would watch him. My two older kids would play in the halls. The department had other female professors, but I was the only one who took her little children to the school. Interestingly, after I started bringing my kids, the men started to bring their children, too.
Evans: How did you get involved in clinical trials?
Wittes: Working with Jerry and the relative betting odds was an introduction to statistical issues related to subgroups and Bayesian analysis.
It was during my Hunter years that I first learned about practical aspects of clinical trials because it was then that I started a bit of consulting. One day, a call came from NIH to apply for the branch chief of the biostatistics research branch at the National Heart, Lung, and Blood Institute. (Many thanks to Bill Friedewald and Curt Furberg who were instrumental in making that call.) I almost didn’t respond because getting that job seemed like such a long shot. But Bob encouraged me to apply and, when the job came through, he went back to the NCI.
When you think of the professional difficulties stemming from my being a woman (and, for me, the problems ended after graduate school), you have to weigh those problems against the fact of a husband’s trailing his wife. I will be forever grateful to Bob for that.
The branch was a terrific place to work. The group was very strong and very productive. We worked very closely with the rest of the institute. We all learned about clinical trials and taught each other about trials, statistics, and medicine. I would have stayed at Hunter forever had the NHLBI not shown up. And I would have stayed at NHLBI forever because the branch was always involved in interesting and important work.
But Bob decided to leave NCI and go to head oncology drug development at what was then Bristol-Myers. I said, “I’m not moving. You stay there for six months. If you like it, then I’ll come.” He liked it and I came, and then Bristol merged with Squibb and he said he didn’t want to live through a merger. So we moved to DC and he returned to the NCI.
Evans: What did you do when you returned to the DC area?
Wittes: At that point, my NHLBI job was filled and I had nothing to do. I got some calls offering me positions, but they were positions for which I never would have given up my tenure at Hunter and now this was eight years later. I was very unhappy. Bob said, “We’ve moved twice in one year. You have a bunch of half-written papers. Just finish them and in the meantime something will show up.”
That seemed like a good idea. Soon after we moved back to DC, before I got much done on the papers, Gerry Sadoff, who was then the head of immunology at Walter Reed, called me with the challenge: “You’re unemployed. How would you like to learn about malaria?” I said, “That sounds interesting,” and that’s how Statistics Collaborative started.
I never expected to start a company or imagined that I was capable of running a business. But once Gerry’s call came, I decided to see if I could do full-time consulting. I only wanted to work on projects that interested me. If I couldn’t do that, I just wouldn’t work anymore. So that was a luxury of being a woman with a husband who can support the family. One can just say, “I’m not going to work if I can’t do what I want.”
Our philosophy at Statistics Collaborative is that we are cautious in choosing projects. The assignment has to involve a somewhat tricky statistical issue, or it has to be related to an intervention that is innovative in some way.
Evans: Do you find that you can still have a very successful business without compromising your interests and values?
Wittes: Yes, but the answer does depend on what you mean by “very successful.” If my goal had been primarily financial, then the route would have been very different. But financial success was not my main goal. Of course, I didn’t want to lose money. I never wanted the company to grow. We started with one person—me. Here it is 23 years later and our head count is about 45, so I’ve actually failed in not growing. But we have succeeded in growing only slowly.
Evans: The statistics profession is sometimes challenged with an image issue. We’re often thought of as technicians and tools, rather than collaborators and strategists. How do you choose the people with whom you work?
Wittes: That’s a great question. Choosing clients is very difficult because you don’t know when you first meet people how they will act. As statisticians, we have to think abstractly, but communicate concretely. Much of our business comes from referrals. When someone new calls, we’ll start a discussion to get a feeling for the person’s values. If the prospective client says, “I’ve been joined at the hip with my statistician and I’m in this new place where I don’t have one and I need one,” then you know you are talking to someone you can work with.
On the other hand, if the person says, “I am talking to a venture capitalist so I need you to give me a p-value that is less than 0.001,” you know this is someone who views you as a tool.
Usually, the differences aren’t so obvious in the beginning. When we are about to start working with someone we don’t know, we like to start with a small project to see how well we communicate. Is this person seeing us as just a technician, or as a partner? When someone says, “I want you to do the analysis, but I don’t want you to write it up,” that’s a bad sign.
Unfortunately, the problem involves not only the people we work with. It is also a problem with statisticians who agree to play being what my husband would call “a statistician from central casting.” They will agree to limit their engagement to a narrow statistical piece without insisting on being broad. I don’t know whether that’s a fault of the self-selection of statisticians or our statistical education system that emphasizes the technical aspects of statistics at the expense of everything else. I realize that schools must insist on technical expertise, but schools also need to teach graduate students that the statistical solutions only represent a part of the problem. And they must do a better job teaching students how to write.
Let me tell you a story. As a first-year graduate student, I took analysis of variance from David Bartholomew, who was a visiting professor from Wales. He assigned us a very complicated homework problem. The set-up was butterflies in cages with a crossover design. I went home, worked really hard, and finally solved the problem (I thought). I proudly handed in my homework, but Bartholomew returned it with only one comment, “What does this have to do with butterflies?”
I realized that I had abstracted the question, solved the math, but had failed to make the answer concrete. I didn’t end up with butterflies in cages. I ended up with a formula and that, I learned, was not the real solution. The lesson has lived with me ever since then. “You have to describe the answer with words that relate to the subject; a formula is not the answer.”
Evans: Yes. I try to get my students and colleagues to view themselves more as strategists than technicians.
Changing gears a little bit, you’re a past president of the Society for Clinical Trials. I had the pleasure of teaching a course with you at one of the society’s annual meetings a few years back. We share an appreciation for clinical trials and the society. What would you see as some of the bigger challenges in clinical trials today and where the science of clinical trials is going? Also, I’ve noticed there are fewer statisticians moving into clinical trials than in the past, at least in the academic setting. We see many people going into Big Data, and I’m hoping we don’t lose too many promising people from the clinical trials arena.
Wittes: You are addressing some really important questions. First of all, I want to say how much fun it was to do that course with you. I still think back to it with warm memories.
Evans: Thank you!
Wittes: I predict there will be a melding of clinical trials and Big Data. I don’t know how or when that’s going to happen, but it has to be true that sometime soon cell therapies and genetic markers and so forth will be much more central than they are now. It’s very clear that clinical trials and treatments are moving in that direction.
While the changes are especially rapid in cancer, they are happening in other fields as well. For example, in cardiology, one can identify genes that make a patient more or less susceptible to a specific drug.
We statisticians did not get into Big Data quickly enough. Our instinctive reactions are in many ways much more cautious than the reactions of people from other disciplines. We should be training people in genetics and in mathematical biology in a way that we didn’t before.
My generation is not an integral part of those areas, but our field must move into these problems. A discipline that stays stagnant will die. You are probably closer to this than I am. Are students less interested in clinical trials than in other fields?
Evans: I think that there is a lot of interest in Big Data generated from rapidly evolving areas of research such as genetics, imaging, and social media. Many of the fundamentals of clinical trials have been around awhile, although many innovative adaptations are still needed because of new problems and new types of data. I agree with you that we need to teach statisticians to understand the medical and biological aspects better.
One worry that I have, particularly in the statistical clinical trial world, is that, in some instances, automated programming has replaced critical thinking. Part of this may be driven by the need for validation and procedures to ensure quality, but I am concerned that these efforts take so much time, leaving less time for understanding the questions and thinking about the most optimal way to address the questions. Do you share this concern?
Wittes: I do share that concern. I also am very worried about comparative effectiveness. People often think that if you have lots of data, you can get answers to clinical effects from observational data. People who are trained outside of statistics seem to be less worried about bias than we are. We’re taught to be very skeptical. I don’t know if that is good or bad, but that’s who we are. I believe it was Bernie Fisher who called statisticians the “terrorists of clinical trials.”
We’re there to look at data, think hard about the experimental design, but then, instead of expressing amazement about the results that have convinced others, we often say why the inference that people are drawing is wrong. It’s not exactly a strategy for winning friends and influencing people. We need to learn to be negative in a very nice way.
Evans: I agree. I think there is a general thinking that more data must be better. But the details regarding how the study was designed and how the data were collected are critical. There may be a part of the research community that would rather have a very large observational study than a modest-sized randomized clinical trial. This of course depends on the details, but without additional details and realizing the importance of randomization, I’ll opt for the modest-sized clinical trial.
We can sometimes get into trouble when we try to use data from sources that were not designed to answer the research question. If having poor quality data leads to uncertainty and misleading conclusions, then what does lots of poor quality data lead to? Indeed we need to be careful.
Wittes: I totally agree.
Evans: Circling back to some of the women’s issues, you described many of the challenges of years ago. But when we look at the younger generation of people in school these days, we’re seeing a lot more women in statistics. Steve Pierson at the ASA had an article in Amstat News last year where he noted there were large increases in undergraduate statistics degrees and women were accounting for a large portion of those degrees.
Do you see a bright future for the next generation of women in statistics? What advice might you have for young women who are considering a career in statistics or who are trying to build one?
Wittes: I do see a bright future, not only in statistics, but in science in general. The doors are open much wider for women than they were in the past. That’s very good news. I worry, however, about men; think of Hanna Rosin’s book The End of Men. In some ways, the pendulum has swung. In many fields, more women are being trained than men are. In many medical schools, more than 50 percent of the students are women.
When one thinks of statistics spanning an axis from the most applied methods (biostatistics or psychometrics) to the least applied (highly theoretical), I would bet that the proportion of women is highest at the applied end and decreases as the field gets more and more theoretical. I’m not sure why that is. Many studies have asked how much of that array by gender is biological and how much is socialization.
Statistics is an ideal field for women—for men, too—it’s very flexible. And if you like to work in a variety of fields, statistics allows that, too. I sometimes describe statisticians as promiscuous; we hop from field to field (or, in clinical trials, from disease to disease). One of our strengths is our ability to see structural similarities across disciplines. In many ways, that is the property of statistics that excited me years ago when I read Moroney’s book.
Evans: What do you think about the increase in undergraduate statistics majors?
Wittes: I am concerned about undergraduates majoring in statistics. An 18-year-old may not be
intellectually mature enough and knowledgeable enough about other fields to be starting in applied statistics. Concentrating too early might not be intellectually healthy.
While I’m not cheering the increase in undergraduate statistics majors, I’m certainly applauding the increase in taking statistics as undergraduates. Even in high school, it’s terrific that the kids are learning to understand probability and how to look at data.
Scott, when did you go into statistics? I know the interviewee is not supposed to ask the interviewer questions, but I am curious.
Evans: My father was a mathematics educator and encouraged me in math. I began to excel at a young age and was pushed ahead a couple of grades in math, such that at the end of my sophomore year in high school, there were no math courses left to take. As a result, during my junior and senior years of high school, I spent half of the day attending college and taking math (and statistics) courses. I majored in math in college and did a master’s degree in pure mathematics in the ’90s, thinking all along that a PhD in mathematics was next. However, when I looked at students in my department who were graduating with PhDs in pure mathematics, I noticed that many were struggling to find jobs. There were many applicants for every position announcement (at least in academia).
But statistics was blossoming with opportunities. I had taken several statistics courses along the way and enjoyed them. I investigated the opportunities in statistics and, after becoming interested in some of the medical applications, I eventually chose to pursue a PhD in biostatistics instead.
I’ve read many of your papers over the years and I’ve not only learned from those papers, but I also enjoyed reading them. You have a writing style that has a flair for creative play on words and telling it like it is. If you look at the titles of many of your papers, they just catch your eye and give you a smile while you’re reading. Some titles include “Missing Inaction: Preventing Missing Outcome Data in Randomized Clinical Trials” and “Stopping a Trial Early—and Then What?” Do you have any advice about effective writing?
Wittes: That’s very lovely to hear. I don’t like separating mathematical thinking from liberal arts thinking. I get very upset when I read something where the word is not right. This sense probably comes from the family I grew up with and from my husband, all of whom care a lot about language. I recently read Kahneman’s Thinking Fast and Slow. His style is so engaging. It just pulls you in and makes you want to read. It’s like reading a novel, except it’s hard science and economics—very serious.
Evans: Do you have any final words for people in statistics or for people thinking of going into statistics?
Wittes: If you understand what Sir Francis Galton meant when he said, “Whenever you can, count,” if you are more interested in the process of inference from data than what the data say, and if you are intrigued by uncertainty, then statistics is a field for you.
Further Reading
Anderson, G. L., M. Limacher, A. R. Assaf, et al. 2004. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy: The Women’s Health Initiative randomized controlled trial. JAMA 291:1701–1712.
Cornfield, J. 1969. The Bayesian outlook and its application. Biometrics 25:617–657.
Findley, A. 1951. The phase rule and its applications, ninth edition. New York: Dover.
Henderson, L. J. 1913. The fitness of the environment. New York: Macmillan.
Kahneman, D. 2011. Thinking fast and slow. New York: Farrar, Straus, and Giroux.
Moroney, M. J. 1956. Facts from figures, 2nd edition. London: Penguin Books.
Pierson, S. 2013. Growing numbers of stats degrees. Amstat News.
Rosin, H. 2012. The end of man: And the rise of women. New York: Riverhead Books.
Rossouw, J. E., G. L. Anderson, R. L. Prentice, et al. 2002. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: Principal results from the Women’s Health Initiative randomized controlled trial. JAMA 288:321–333.
Wittes, J., E. Barrett-Connor, E. Braunwald, et al. 2007. Monitoring the randomized trials of the Women’s Health Initiative: The experience of the Data and Safety Monitoring Board (PDF download). Clinical Trials 4:218–234.
Wittes, J. 2009. Missing inaction: Preventing missing outcome data in randomized clinical trials.
J Biopharmaceutical Stat 19:957–968.
Wittes, J. 2012. Stopping a trial early—and then what? Clin Trials 9:714–720.
About the Author
Scott Evans is the executive editor of CHANCE and a senior research scientist at Harvard University, where he is the director of the Statistical and Data Management Center for the Antibacterial Resistance Leadership Group and teaches clinical trials.