Interview with Melissa S. Anderson
Interview conducted in 1995
"We have also identified what we call 'questionable research practices' that don't violate any specific research rule, but can clearly lead to serious problems."
Melissa S. Anderson studies mathematics at St. Olaf College in Minnesota and received a masters degree at the University of Iowa. She completed a Ph.D. in education and is now on the faculty at the University of Minnesota. She has consulted with the Acadia Institute on research practices and ethics and her research focuses on trends and changes in higher education.
Aczel: Where did you go to school and what type of statistics courses did you take? What did you think of them?
Anderson: I did my undergraduate work at St. Olaf College in mathematics, my masters at the University of Iowa in mathematics, and my Ph.D. at the University of Minnesota in higher education. I took several courses in probability and mathematical statistics from Dick Kleber at St. Olaf. He's legendary at St. Olaf and made everything interesting. Everyone remembers the old mechanical slot machine he brings to class to illustrate statistical principles.
Aczel: What did you do after you graduated from college?
Anderson: I went straight into graduate school after college and then taught in the math department at St. Olaf for five years. Then I decided to go back to graduate school in a new area, higher education. I'm now a faculty member at the University of Minnesota in higher education. My research here is focused on graduate students, faculty, academic departments, and institutions of higher education. I teach courses on administration in colleges and universities, the history of higher education, the development of academic programs, educational institutions as organizations, policy research, and so on. I continue to study faculty demography in departments (the faculty experience question). I am also involved in the Acadia Institute's Project on Professional Values and Ethical Issues in the Graduate Education of Scientists and Engineers. We have been looking at, among other things, faculty and doctoral students' experiences with academic misconduct in their departments, their values with respect to research and education, and their views of the climates of their departments as places to work. We have been using a new statistical approach, hierarchical linear modeling, as developed by Tony Bryk and Steve Raudenbush, to look at how departments affect the experiences of their students and faculty. It's always exciting and frustrating to explore new statistical methods - it always reminds you how much sympathy students deserve as they struggle with statistical ideas for the first time.
Aczel: How did you first get interested in statistics?
Anderson: I always enjoyed stats classes, but I really got interested in statistics when I started work on my dissertation and had to learn a lot about longitudinal analysis on my own. I wanted to see how the range of faculty experience in a department is related to the department's funding over time. Building models of change over time is exciting, because so much research deals only with what is going on at a particular point in time.
Aczel: What is the mission of the Acadia Institute?
Anderson: It provides research, consulting, and education programs. The Acadia Institute is a nonprofit research institute located in Bar Harbor, Maine. It focuses on scientific and medical issues with particular attention to ethics, research, and graduate education.
Aczel: What are the particular issues Acadia is focused on now? Are there any unique studies or concerns as related especially to statistics?
Anderson: Besides the graduate education project in which I've been involved, Acadia is doing work on the federal system for the protection of research subjects and on undergraduate medical education. It is also developing a network for bioethics.
Aczel: We've seen some ethical issues in the medical research field, such as falsification of data, come up in news lately. What about academic research?
Anderson: We did a survey of 4,000 faculty and doctoral students nationwide in four fields: chemistry, civil engineering, microbiology, and sociology. We asked them if they knew of people in their own departments who had falsified research data. Six percent of the faculty knew of other faculty who had done so, but thirteen percent of the students knew of faculty who had falsified data. Also, about eight percent of the faculty had seen students falsifying data, whereas twice that proportion of students had seen other students doing that. Given the seriousness of this offense, these numbers are really disturbingly high.
Aczel: Are there real ethical issues statistics professionals should be concerned about, or is it just a question of "bad" versus "good" statistics?
Anderson: All researchers must be aware of the ethical standards and practices that apply to their fields. Some of these standards are formal (for example: don't plagiarize, don't fabricate data, don't misuse research funds, don't cause harm to research participants). But we have also identified what we call "questionable research practices" that don't violate any specific research rule, but can clearly lead to serious problems. This category includes, for example, looking the other way when you know someone you work with is using data inappropriately, and skimming over data or statistical results that contradict what you'd like to prove.
Aczel: For students today, what issues in statistics and research will they likely hear about in the future?
Anderson: In my own corner of higher education, there's a lot of excitement about statistical approaches that let us examine changes and trends over time, longitudinally. We're also using new ways of looking at relationships across levels; for instance, we're studying students who are in departments that are in colleges - all these levels complicate the kinds of statistics that are needed. New statistical models that will make this kind of study reasonably easy are just now emerging. It's really fun to explore these new approaches.
The important part about statistics, or any field, is that new approaches usually address deficiencies in what we've understood up to this point. You can look at simple relationships between, say, a student's financial aid and his or her chances of completing college. But the underlying issues, like most of life, are very complicated. If you look at lots and lots of other factors in this relationship (which requires a very complicated model), you see that the simple relationships you thought were clear are actually inaccurate. The point of statistics is to figure out what relationships are really out there in reality, in life - not for any one individual, but for big groups of people. The better we can understand that reality, the better our chances of making good decisions and good policies that affect those broad groups of people.
Aczel: Do you have advice or suggestions to students in their first (and possible only) statistics course?
Anderson: Ask questions!! Question every step that you don't understand. Question the value and use of every new approach. Ask good questions about how the statistics you are learning are applied. Ask yourself questions to see if you can anticipate what will come next. Ask each other questions to get a different spin on the material. Answers are all well and good, but it's the questions that really move you ahead.
Statistics can be rough. I know a former college president, now in his 70s, who still wakes up in a panic in the middle of the night, thinking that he's about to face his statistics final exam. I think that asking questions give students more control over what they are learning and eases the pressure quite a bit.
Aczel: What could educators do better for statistics?
Anderson: Make sure that students know how critical statistics are to virtually any responsible position in our society. Incorrect statistical results or inappropriately interpreted results can have devastating consequences.
