ON STATISTICS PROGRAMS

Robert V. Hogg
University of Iowa

Newsletter for the Section on Statistical Education
Volume 3, Number 2 (Summer 1997)


  1. Background:

    As of June, 1997, I will have been at the University of Iowa for 50 years. During the first three years of those fifty, I served as a graduate assistant at what was then called the State University of Iowa (SUI). With the name Hogg, people often observed that they could see why I was attracted to SUI, due to that pig called "sooo-eee". However, with my Ph.D. in hand, I became an Assistant Professor of Mathematics in 1950. Later in 1965 we started the Department of Statistics and Actuarial Science. While the Actuarial Science in the name was added later, we always included the actuaries in this new department, and that has been a very satisfactory "marriage".

    In addition to being the Executive Officer of Iowa's department for 19 years, I have visited many other Departments of Statistics. I have learned a few things, some from the good and some from the evil, namely the various mistakes, many of those being mine. Finally, during the period January-May, 1997, I visited statistics units at 14 universities; some of these were in Industrial Engineering, Business Colleges, and Mathematics, but most were Departments of Statistics. While these observations are fresh in my mind, I decided to write two reports: the first dealing with Statistics Programs and a second concerning Continuous Quality Improvement in Higher Education, which will appear elsewhere.

    Before focusing on statistics programs, let me address a feeling that I have about statisticians and the statistical profession. Frankly, I do not find enough cooperation among the members of our community. Are we really supportive and flexible enough when the opportunities arise to help other statisticians? These might involve visits to other universities, special research or teaching opportunities, or evaluations of grant proposals. Do we really reach out to help? I guess that I want us to be a family of statisticians working together for the good of the professions. We should want to share information, benchmark other programs, and recruit more young people as professional statisticians. Moreover, we should do much more to sell statistics to the general public. Statisticians can be collaborator (even leaders) on major projects, and yet very few have any notion that this is possible. Instead many of us are faced with a reward system that almost forces us to be loners; certainly the tenure system is laden with fear.

    The American Statistical Association (ASA) should be, and is to some extent, addressing some of these concerns. However, we need much more substantial efforts at a national level to achieve the necessary progress. A short session involving departmental chairs and heads at our annual meeting is not enough to discuss seriously major concerns about the directions of our profession. Let us do as many other professions do and get key people together at least once per year and brainstorm about appropriate actions that will benefit the profession. Statistics has been fairly strong in the past and super for me personally for 50 years; I only hope that the young people of today can have that same viable option that I had fifty years ago.

  2. Present Graduate Programs:

    Various statisticians can come up with different ideas about the nature of statistics, but I think most of us would agree that the following is close to what we want to do in the practice of statistics: (a) Create measures for problems under consideration; (b) Collect data through surveys, experimentation, or observations recognizing that some uncertainty exists in these data; (c) Analyze the data and provide information, again with an element of uncertainty; (d) Prepare a report with recommendations, beginning with a brief executive summary (that is, KISS: Keep It Simple Statistician).

    It has been my observation that we spend most of the time in our academic programs on point (c). The usual Applied M.S. program is something like this: 2 or 3 courses in probability and mathematical statistics; 3 or 4 courses in regression, design and analysis of experiments, and multivariate analysis. These courses include response surfaces and computing (that is, the use of some statistical software); 3-5 electives from time series, nonparametric statistics, data analysis, sampling, statistical quality control, consulting, and categorical data. A more theoretical M.S. program for Ph.D. students would contain some mathematical analysis and an advanced probability course or two in place of some of the electives.

    Most universities with statistics departments have created some sort of statistical consulting center; such a center provides valuable experience for the students. I believe that these centers should be more aggressive and provide experiences for more students than they do at present. Also most statistics departments seemed to have reasonable computing facilities, but these must be improved continuously as advances are made in technology.

    It is clear that not enough is done to recruit students (particularly Americans) to our profession. While this might be a job for ASA, each of our departments can contribute some to this effort by visiting nearby undergraduate programs in math (or even high schools). Each such effort helps, even though most of us would be thinking primarily of our own department. It certainly must concern all of us to see the closing of an occasional department of statistics in this country. We believe that statistics is important, and we must attract a sufficient number of students. In my opinion, most of our programs are not flexible and modern enough. I will say more about this later because if we do not change, we will see more such closings.

  3. Systems Approach:

    The department should decide on its mission, purpose, aim, goals etc., and have a team to design a curriculum to reflect these (see Section 7 on Curriculum Review). After consulting appropriate persons (students, alums, businesses, other faculty in and out of department, etc.) create a "core" for each desirable program.

    In particular, a Board of Advisors consisting of alumni and other influential friends might be most worthwhile. To maximize any system we must recognize that we are dealing with many interdependent parts, and we can not just try to maximize each of them. We really want to create a community of scholars, working together and recognizing that all of us do not have the same strengths or interests. Hence these cores reflect what we think best for the students in each of our programs. Some courses (core and elective) would possibly be team taught.

    My guess is that a first-year statistics core in graduate school will consist of some studies that deal with theoretical, applied, and computational skills. Possibly we have not stressed the latter enough in the past, but clearly we must consider the present and future technologies and take advantage of them. These include knowledge of excellent statistical software, spreadsheets, managing data bases, and data mining, in addition to being more concerned about the quality of these data bases. Others (computer science, electrical engineering, business, etc.) will be (or are) teaching these if we are not interested. With large data sets, nonparametric methods can be used to determine the "middles." There is still a concern about the variation (skewed, heavy tailed) and statisticians can help, if we will, about predictions concerning future observations.

    Certainly no one course should be "owned" by one professor if others are capable of teaching it. We might look forward to the day in which an expert in some subject who is at another university and teaches his/her specialty to those at other universities; this possibility is closer than most of us believe. In this regard, I find that we require too many courses for the Ph.D. degree. Beyond the Casella and Berger level, we need a good theory sequence, a Linear Model/Multivariate sequence, and a strong probability sequence. After that students can take electives, maybe given through Topics, possibly taught by one of these outside experts. These courses might include topics such as nonparametric regression (estimation and graphics in general), spatial statistics, computer intensive methods (particularly with Bayesian methods and resampling), empirical processes, non-linear dynamics, stochastic differential equations, and sequential methods (including meta analysis).

  4. The Statistical Community:

    (a) Senior faculty should be mentors for junior faculty and graduate students in research and in teaching. Graduate students should, at least once per semester, receive some report on their progress, but this would be better given on a continuous basis.

    (b) Advanced graduate students should help the beginning students. It would be worthwhile to have a weekly seminar for all graduate students. Three or four students would report each week on topics appropriate for the levels of the students in that program. The faculty advisor would assign the topics (possibly with help of students) and require attendance of all students in the program. The graduate students would get to know each other so as to help one another and hopefully create a little "esprit de corps." Such a seminar would also help improve the "people skills" of the graduate students; this would also be true with involvement in the consulting service.

    (c) Faculty members should discuss their experiences in various courses with other instructors, particularly with those who follow teaching the same courses. It is important that we agree on topics in one course that is a prerequisite for another; otherwise the instructor of the second course has big problems.

    (d) We should ask for feedback from students (minute papers, punctuated lectures, quality teams reporting each week) and give them feedback on the feedback. Students cannot tell us what to teach, but they know when they are bored or confused. I am convinced that all of us want to be better teachers, and we should discuss among community members how to improve. As an example, would it be helpful to put notes on the web? The students would like this, but attendance might be worse than it is now in large lectures. (Note: I have found that providing students with solutions of quiz and test questions immediately afterwards is beneficial.) In general, there should be more interaction among students and their instructors. (See Hogg's "Continuous Quality Improvement in Higher Education" for more suggestions.)

    (e) Leaders should be helpful in facilitating the professional development of others. People want to feel good about themselves and their efforts; so real effective leaders should try to end discussions (some can be painful) on some kind of positive note. It never hurts to ask about a spouse or the children; such a personal interest lets the other person know that you care about him or her and his/her family. This is important (and I'm not always certain that I was real good at this in the past; I must improve).

    (f) To address some of these items (and others like the reward structure), an occasional retreat of the faculty might be very valuable. These (as well as other meetings) can be overdone, but sometimes they are needed to discuss seriously the goals of the unit and the best ways of achieving them. Maybe these retreats, usually held most successfully off- campus, would make department members feel more like being on one team that is an important part of the university.

    (g) The present reward and tenure structure is such that many tend to be more loyal to the profession than to the university. More should be done to interact with others on campus, possibly collaborating with faculty in other fields. Such interaction would make us feel as if we belong to the university community.

  5. Partnerships in the Extended Community:

    We must search for these partners. As statisticians we have a certain advantage in cross-disciplinary activities as most researchers will collect data and will need these analyzed to get the maximum amount of information from them. These partners can be from our own campus, possibly resulting in joint research (grants, contracts, etc.) for faculty or cross- disciplinary theses (co-majors) for our students. Off-campus partnerships can lead to consulting, internships, or projects involving some unstructured problems. Often if these are close enough to campus, M.S. or Ph.D. theses could result from these involvements in substantial problems. Certainly stronger and more aggressive statistical consulting services with strong faculty involvement will help promote some of these partnerships. And these outside involvements certainly can not hurt, but almost always improve, the people skills.

    In this regard, I wish that some statisticians would be entrepreneurs. We must sell the value of statistical thinking to others. We must explain to others the power of statistics as being very supportive to good research involving the collection of data. Often examples and case studies could be useful in such situations, encouraging students in other fields to take more statistics. Of course, some of our Ph.D. students should be involved in cross-disciplinary research, possibly through co- majors with another area.

    Then too we can convince others that statisticians can help in their programs. For example, if a Business College is preparing students as Quality Managers with courses in human relationships, planning, and budgeting, appropriate statistical methods could also be most useful in such a program. The Japanese recognized the importance of the technical aspects in this area of quality improvement and used it. Moreover, there are many areas, in addition to Business, that need statistical help. Certainly joint appointments in these areas would be worthwhile in various situations. It has always been amazing to me why joint appointments are much more successful at some universities than others. Maybe it is due to different cultures or leadership.

    Of course, situations at some universities could call for efforts larger than helping a few individual programs. It could be that certain colleges (Medicine, Business, Education) have very little statistical help for their research. Depending upon the situation, the organization of a Statistical Institute (or Center) might be very appropriate. This might be difficult to sell such a unit, but then some of us should be entrepreneurs. Give it a try.

    Often a simple way to wave the statistical flag when there are enough statisticians in a Mathematics Department, say, is simply to rename the department as Mathematics and Statistics. Sometimes certain mathematicians object to this change. It is difficult for me to see why this is opposed as such a move would clearly help to recruit students to that department. Many departments have done this very successfully. As a matter of fact, it might be extremely helpful to work with the mathematicians and introduce a course in "Introduction to Mathematical Sciences" so that those with mathematical ability can see the possible options for them in the future.

  6. Service Courses:

    In our profession, we have the opportunity to teach many service courses and we should try to find additional ones when appropriate. Since they are often our "bread and butter" activity (for graduate student support), we must try to improve them on a continuous basis. In particular, we must address how best to deal with the large lecture courses. None of us is really happy with the present situation. Yet we must recognize that we can not use TAs to teach smaller sections because this would defeat the university's mission to have more faculty in those freshman/sophomore courses. There are people in the profession that believe we should try to emphasize "statistical thinking" rather than recording lots of statistical techniques. Many instructors find that student projects truly help in this thinking. I'm inclined to agree with them, but I recognize that the big majority of students taking those courses simply want a grade (hopefully A or B) to satisfy a requirement rather than learn a little statistics. Maybe we should be satisfied if a few (perhaps the top 25%) understand our message and thus teach to them. Then tell the others how to get "that grade" by being able to "plug in" a few numbers in some formula. Maybe by doing the latter, the students will get some idea about the error structure of a statistic and thus understand why statisticians always put a "plus or minus" after our estimates. Nevertheless, I do believe that we should address the problem of large lectures (they are here to stay) and maybe a little brainstorming or benchmarking will help improve the situation somewhat. For example, in teaching statistics (as with mathematics and languages) the students can not miss the first 4 weeks and expect to pick up immediately as they might in history. This might suggest that we create modules so that different students can proceed at different speeds. After all, all do not learn in the same way. Our smaller and somewhat more advanced service courses in statistical methods and mathematical statistics are in better shape, but we should continue to check with our "customers" (other departments) to make certain that we are doing the best possible job.

  7. Curriculum Review:

    I believe that in most statistics departments the curriculum has developed in a somewhat ad hoc fashion, and it is revised from time to time by making minor modifications of the previous plan. Most often the curriculum does not represent the department's goals, even in cases in which these are spelled out. Accordingly I would urge each statistics department, possibly in cooperation with other statistics departments, to assess seriously its curriculum. To help us do this, I have modified an outline that the Southeastern University and College Coalition for Engineering Education (SUCCEED) has created for engineering departments. This modification was made from an outline given in SUCCEED's Executive Summary, and its full report has not yet been finalized.

    (a) Strategic Planning. The members of the department should meet and discuss seriously the situation. Often this can best be done in a one-day retreat away from campus so that distractions such as phone calls can be eliminated. Hopefully the faculty can agree on such things as the purpose, mission, goals, and aims of the department. Does the present curriculum satisfy these? If not, some general principles could possibly be agreed upon and consideration given to guiding principles of the revision. Such a revision might consist of minor adjustment to the present program; but, on the other hand, it might involve a major re-engineering that, for example, might involve a team-taught core program for first-year graduate students (and possibly some very good undergraduate students). However, at the end of this phase, a decision would be made whether or not there is support to continue the consideration of a curriculum revision.

    (b) Preparation. If the decision is to continue, a Curriculum Design Team (CDT) should be formed. While input from junior faculty members, as well as others, is important, it is probably best not to have untenured faculty serving on the CDT. In analyzing the existing curriculum, it would be very important to get feedback from recent alumni in order to find out the present program's good and bad features. The CDT should also benchmark other existing programs that seem to be cutting edge of the profession. With as much background as possible on new and old areas, we can progress to designing a new curriculum.

    (c) The Design of the New Curriculum. With the information found in (b), some consensus process should be established to help the faculty select a new curriculum which reflects the goals of the department. As might be expected, "turf battles" might be fought, but the CDT and chair of the department would need to resolve these as well as possible. Once this is done course-specific issues should be addressed. At the conclusion of this work, an overall structure of the new curriculum will result with the identification of its component parts and the division of the subject material into course-sized segments.

    (d) Beginning of the New Curriculum. Once agreed upon, the chair and the faculty will begin to implement the new curriculum, with much depending upon the timing and funding of this new venture. Assuming the funding is available, the CDT will lay out a schedule that will implement the new program within two years. It should be understood, as the new curriculum is taught, that it should be continuously assessed and improved.

  8. Conclusion:

    I am certain that I have not mentioned everything that should be looked at as we consider necessary changes. However, if we look at our department as a system, we might be able to utilize our members more effectively, even giving some larger teaching loads depending upon the abilities involved. Of course, different assignments would need to be taken into account in rewarding those individuals. We do need improvement, however, and some might enjoy reading my "Continuous Quality Improvement for Higher Education" along with this report;. I am hopeful for the future as we make appropriate changes.

For more information contact:
Bob Hogg
Dept. of Statistics & Actuarial Science
University of Iowa
Iowa City IA 52242
(319) 335-0824
bhogg@stat.uiowa.edu


Return to V3 N2 Contents
Return to Newsletter Home Page