Home » A Statistician's View

In Response to ‘Statistics as a Science, Not an Art: The Way to Survive in Data Science’

1 April 2015 909 views No Comment

Joseph E. Bauer is the director of survey research and evaluation at the Statistics and Evaluation Center (SEC) of the American Cancer Society — Corporate Center in Atlanta, Georgia.

This was an article that I considered spot on! It should get everyone to think about our profession and what we might be able to do as individual statisticians.

The beginning of the piece starts with advice the author’s father gave him: The most important thing about solving a problem is to formulate it accurately. I received similar advice from my father to “measure twice and cut once”—which, for me, translated into spending the time on the study design, research question(s) and hypotheses, and theory before deciding on a statistical approach and implementing the study, which is all about formulating the problem accurately. When one takes the time to do this, and this happens when you talk with collaborators (who are the nonstatisticians), you increase everyone’s appreciation and understanding of the problem, as well as the value that you (the statistician) bring to the table.

stat_view

Do all statisticians blithely select a ‘convenient’ statistical model (akin to ‘if you have a hammer, then all problems become nails to be whacked’)? Well, probably not all, but undoubtedly, many do. (If one is good with survival models, then the tendency is that every problem tends to be formulated with a survival model solution as an almost ‘canned solution’ [hence ‘convenient’]).

The author’s larger point is that if many statisticians are doing this, then our profession is in trouble! Maybe it is exactly this kind of behavior that has precipitated a ‘de-professionalization’ such that collaborators orient toward statisticians as ‘technicians’ who, as the author says, “the collaborators can steer to get their scientific results published.” (How often have we heard “We just need the p-value to be less than .05”?) How many statisticians just provide that?

Of course, we have a bit of a chicken and egg problem. Is it the behavior of the statistician that causes these issues, or is this what happens when collaborators come to us at the last minute, say on grant submissions? In that scenario, they want us to add a statistical design and analysis plan on half a page (because that’s all the space left in the proposal) by tomorrow morning, without the benefit of getting to read and think about the rest of the proposal. The causal dynamics probably happen from both directions. In any event, we need to think actively about what we can do to increase professionalism and being treated as professionals.

It seems to me that if this phenomenon is happening on a wide scale, this is a violation of consulting 101. Formulating the problem accurately is the goal, which means the necessary work (the hard work) is talking with the collaborators and listening to them (and vice versa) about how the data were generated and what their thinking is about various parameters, etc. How to get that dialogue started and, ideally, early in the project planning cycle is the challenge.

As the author says, “We will open up a new world to our collaborators to actually being able to generate questions our collaborators had no idea they were even allowed to impose.” We have all likely had consultations that run the gamut. The question becomes how do we get more/most of our consultations to be of the type in which “we are working together with our collaborators to correctly formulate the problem”?

I fully agree with the author when he says, “This field is open for all to contribute to, and the truth is that anybody who honestly formulates the estimation problem and cares about learning the answer to the scientific question(s) of interest will end up having to learn about these approaches and can make important contributions to our field [i.e., statistics].” However, I would add the obverse is true as well: Statisticians can contribute to disciplinary fields outside their specific academic specialty. For myself, I have learned so much from other disciplinary fields and the problems my collaborators face, which is what makes consulting fun and interesting.

We definitely need to be part of that scientific/project team solving real-world problems together. However, there are many challenges here. The quandary is that there are many collaborators who want to ‘objectify statisticians’ and place us in categories (i.e., geek or nerd)—and some of us wear those labels proudly. Being characterized like this tends to reflect myopic expectations—that all we do is generate some statistical formulas or graphs. Many collaborators perceive that statisticians have no disciplinary knowledge outside of that stereotype and, in the extreme, that we are not even a professional discipline, but simply technicians. This stereotype further implies that statisticians do not have any desire to interact on things that are not ‘statistical’—and maybe even that we have no wish to be co-authors on manuscripts or co-investigators on grants. These conditions may, in fact, precipitate the conditions whereby statisticians gravitate toward the ‘convenient’ solution and thus behave in a way that leads to further de-professionalization. How do we break this self-defeating cycle?

I think the author gets us all to think more deeply about our profession. It is a clarion call about statisticians and our profession being marginalized (and how many of us may be contributing to our own demise as individuals and the marginalization of our profession). We need to ask ourselves, have we abandoned theory and have we abandoned trying to have that more robust discussion with our collaborators?

1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)
Loading...

Comments are closed.