Skip to content

Social scientists who use medical analogies to explain causal inference are, I think, implicitly trying to borrow some of the scientific and cultural authority of that field for our own purposes

I’m sorry I don’t have any new zombie papers in time for Halloween. Instead I’d like to be a little monster by reproducing a mini-rant from this article on experimental reasoning in social science:

I will restrict my discussion to social science examples. Social scientists are often tempted to illustrate their ideas with examples from medical research. When it comes to medicine, though, we are, with rare exceptions, at best ignorant laypersons (in my case, not even reaching that level), and it is my impression that by reaching for medical analogies we are implicitly trying to borrow some of the scientific and cultural authority of that field for our own purposes. Evidence-based medicine is the subject of a large literature of its own (see, for example, Lau, Ioannidis, and Schmid, 1998).


  1. Melvin Livingston says:

    This seems a reasonable critique. However, coming from an epidemiology background it seems perfectly natural to do. All of our study design and causal inference education is firmly grounded in the evidenced based medicine literature. However, many epidemiologists fall somewhere between medical researchers and social scientists (at least in practice). While doing our social science work, the more concrete medical analogies are the easiest thing to reach for when trying to explain things to a wider audience. The social and behavioral research we do always feels rather segmented. Everyone has their specialty, but not everyone knows enough about your specialty to follow along. Luckily, most epidemiologists have plenty of experience with the medical literature (at least while in grad school); it acts as lingua franca for us.

  2. Jeff Lax says:

    I do this all the time — I use medical treatment examples because most of us now what the right setup is for testing something in that domain. Not that most medical research reaches the level of a good undergraduate honors thesis in political science. But the ideal is clear. I’m borrowing the analogy to their ideal, not their practice. Reading a random journal article in medicine can make even empirical judicial politics look good.

    • Andrew says:


      I see what you’re saying, but the flip side is that, by comparing to an ideal, you’re idealizing and disengaging the method from the students’ substantive interest. If, instead, you use a get-out-the-vote experiment or a public opinion experiment as your example, you give your students (and yourself!) a chance to understand the theoretical assumptions in a context where you can better evaluate them.

      • Fernando says:

        Agree 100%.

        The medical analogy is useful as a first introduction to the topic of experiments.

        Thereafter you want to connect the method to the audience’s substantive knowledge so they can judge for themselves.

        • Jeff Lax says:

          The students’ substantive interest may be many things (though probably not get out the vote experiments in most cases), but their most likely exposure to the experimental ideal is indeed in terms of med examples. Fernando is 100% right. :)

          THEN the point is to see whether the assumptions/setup from the ideal match the case they care about.

  3. Fernando says:

    A simpler interpretation is the deliberate use of a preexisting schema to clarify ideas. And in this regard medicine has some advantages:

    Unlike physics or agriculture medicine deals with human subjects, which is more relevant to social scientists.

    Unlike psychology medicine also deals in concepts that are better defined and easier to measure (e.g. a specific vaccine, disease outcome).

    And medicine is familiar to most people (e.g. aspirin, headache).

    Thus, the virtue of medial analogies is that they lie somewhere between clean physical experiments, that often eschew randomization, and messy psychological experiments, that can seldom be replicated let alone understood.

  4. JSB says:

    Perhaps researchers are attempting to gain authority though citational osmosis, but I wonder how much is actually filtered through. Medical research has strong and diverse suite of critics. On the whole though, I think familiarity, as Fernando points out, is major strength appeal to non-medical researchers. In particular, the struggle to connect large scale patterns to treatment of individuals (a real challenge for evidence based medicine) is something that a number of disciplines can draw from.

  5. Matt says:

    I suspect people use medical examples because bad statistical practice literally kills people (and good practice keeps people alive). The line of causality from statistical methods to life-and-death results is harder to draw from most social-science research.

  6. Robert says:

    By the way: In this introduction to statistics for high energy physics, they also use a medical example (actually to explain Bayes’ theorem). (First set of slides, p.11) Maybe it’s not about borrowing authority from medicine, but it seems to me, that medical examples are easy to understand.

  7. […] Three good posts from statistician Andrew Gelman: 1) health disparities are associated with low life expectancy; 2) researcher “degrees of freedom” (ie, how many ways can you fiddle with the data to get the result you want; and 3) social scientists should stop borrowing medical terminology. […]

  8. […] Medical Analogies in the Social Sciences Andrew Gelman writes: […]