A degree in statistics?
Some statistical consultants claim that to be a statistician, a person must have a degree in statistics. Thankfully, such a “degree bias” did not deter people like R. A. Fisher (astronomy), Karl Pearson (biology), William Gossett (chemistry), W. Edwards Deming (mathematical physics), John Tukey (mathematics), and Lee Cronbach (psychology) from building the foundations of modern statistics. This bias also implies that only through a degree in statistics is a person able to understand and “do” statistics. This seems like an unnecessarily discouraging stance to adopt in trying to help people like doctoral students and researchers understand statistics. It is also the case that some of the most significant advances in statistics have been instigated or led by non-statisticians trying to address problems in applied fields such as psychology (e.g., factor analysis, testing theory), social research (e.g., structural equation modeling), marketing (e.g., multidimensional scaling, conjoint analysis), and management science (e.g., linear programming).
As a final nail in the “real statistician’s” coffin, I need only ask how successful you have been in learning statistics from people with degrees in statistics? The best teachers of statistics I have had were people who had the knack of conveying the various concepts and methods in simple visual ways that avoided mathematical derivations. Once you grasp a statistical method in conceptual terms, it is easy to move to an understanding of the results the method produces, the numerical operations needed to compute the results, and ultimately, the mathematics of the method. Most mathematical statisticians assume that the subject is best learned the other way around: teach the mathematics first, and then trust that the conceptual understanding will follow. This may work for the tiny fraction of statistics students who are mathematically gifted, but for the rest of the students (which I suspect exceeds 98% of the total), starting with visualizable concepts works best. I have taught statistics and research design for many years, and am very good at equipping my clients with the degree of understanding they seek to acquire of the methods I am applying for them.
Membership in professional organizations?
So if a degree in statistics is not the necessary hallmark of a competent statistical consultant, what other criteria should guide you in your search for someone to assist you with your statistical requirements? Some say that any statistical consultant should be a member of the American Statistical Association and its Statistical Consulting professional interest section. The ASA is a great organization and provides invaluable services for the communities of academic and professional statisticians and methodologists. However, membership in the ASA is open to anyone with $160 per year to spend, which allows free membership in up to three professional interest sections. Unlike the American Psychological Association, which requires proof of a Ph.D. or comparable doctorate in psychology from an accredited university, the ASA requires no such proof. Thus, ASA membership carries with it no proof of having passed the academic requirements for the teaching, research, or practice of statistics.
The same lack of value as a basis for screening potential statistical consultants also applies to any other groups and organizations which establish no measure of education or experience as requirements for membership. This includes such affiliations as the SAS Affiliate Program and other commercially-based networks, networking groups such as Linked-In, and other statistical interest groups.
Best indicators of a statistical consultant's competence and honesty
Now that I've addressed some of the fallacies about appropriate criteria for selecting a statistical consultant, let me offer some "shopping tips" derived from over 30 years of working in statistics and with other statisticians regarding the choice of a statistical consultant.
- Formal graduate-level coursework in statistics
- Published peer-reviewed articles reporting the results of statistical analyses
- A record of working in positions requiring the use of statistical analyses of the type you are seeking assistance with
- Experience in teaching statistics (better able to explain to you what is needed and why)
- Number and types of clients served in the past
- Testimonials from past clients (if not available on website, ask for one or two former clients who will serve as references)
- If your needs are specific to a particular software package, make sure the statistical consultant either has it or can obtain it within your time window, and that he/she is proficient in its use.
- Ensure that the statistical consultant is proficient with the methodology that you have learned from independent sources (e.g., professors, colleagues) is required or optimal for analyzing your data.
A Final Criterion: Your Statistical Consultant Should be a Good Problem-solver
Many, maybe even most, of my clients come to me with either poorly formed hypotheses or research questions which can’t be addressed statistically, or with no idea of how to use their data to test their hypotheses/research questions, or with both of these problems. The most important services I or any other statistical consultant can provide clients are:
- ensuring that hypotheses/research questions are testable using statistics;
- designing analyses that sufficiently account for the complexities in the data to produce defensible statistical results;
- finding ways to squeeze the maximum amount of information out of any given data set.
Providing the above services requires more than just a technical knowledge of statistics--it requires the consultant to be a creative problem-solver. Just about every client has something unique about their data or the questions they seek to have answered. Meeting these needs requires solutions customized to that client, often involving novel applications and combinations of methods. You can’t come up with such solutions if you just “go by the book”. You need to understand what’s in the book so well that you can go beyond it.
One of my foremost strengths is that I am the quintessential problem-solver. Somehow, I manage to immediately perceive the fundamental elements of any research design and data analysis challenge. The solution to how to best approach a problem then becomes immediately apparent to me. So far, this talent has never failed me: in over 30 years of assisting students, researchers, and business clients, I have never failed to come up with an effective solution to their research design and analysis problems.