Statistics teaching has been a massive educational and ergonomic failure.

by reestheskin on 14/10/2015

Comments are disabled

My intercalated degree (1980) was centred around epidemiology, and analysis of what was then a large dataset. So, I had to learn some FORTRAN and file editing on an IBM360 / find my way around a MTS system, all in the days of punchcards, and line printers, with only one ugly green on black monitor to share across a unit. Mostly, I wasted large amounts of print out and got used to retrieving my batch processed requests early morning, with the comment ‘run aborted’ , ‘system error’ etc. I had to wait another 24 hours to find out if I had corrected the errors in my programs. I remember the first time I looked at the GLIM manuals, wondering if I was attempting to read them upside down.

Years later, my interest in statistics resurfaced, mainly as a response to the banality of much of the EBM crowd with their NNT and their apparent lack of understanding of what I disparagingly call ‘probability management’. Most EBM merchants are frustrated chartered accountants. Real science is involved with understanding reality by creating models of how the world works: if you don’t do that, you are not doing science, but rather the the D of R+D, or ‘technology assessment’. The theory of how you should do the latter is a proper academic pursuit, but using these ‘products’ is not a matter for the academy — although it is important for many businesses or professions. Using what others invented is what doctors do in the clinic, but it does not count as research, but as honest professional toil (and very valuable work, at that, compared with say much business).

But statistics is hard, and frequently counterintuitive. We do not teach it well to medical students and for all the mantra about doctors’ communication skills, fluency with statistics is a core medical skill, and in many situations, the key communication skill doctors should possess. If you want your students to communicate well, do not stray far without mentioning binomial, poisson, Bayes etc.

What follows is a comment from Sander Greenland, on Deborah Mayo‘s excellent siteError Statistics . I do not know Greenland (although we have emailed each other in the distant past), but I think he is somebody who is always worth listening too. There are a couple of points he makes that chime with me, and they relate to both teaching and the ‘crisis’ in much medical (and scientific research). He writes:

“My view is that stats in soft sciences (medicine, health, social sciences among others) has been a massive educational and ergonomic failure, often self-blinded to the limits on time and capabilities of most teachers, students, and users. I suspect the reason may be that modern stats was developed and promulgated by a British elite whose students and colleagues were selected by their system to be the very best and brightest, a tiny fraction of a percent of the population. Furthermore, it was developed for fields where sequences of experiments leading to unambiguous answers could be carried out relatively quickly (over several years at most, not decades) so that the most serious errors could be detected and controlled, not left as part of the uncertainty surrounding the topic.”

First, we have to accept that we have failed. Second, all too often we are ‘self-blinded to the limits on time and capabilities of most teachers, students, and users’. This is a widespread problem in the undergraduate medical curriculum, more generally. Students must be researchers, teachers, scholars etc. All too often this is one giant GMC inspired delusion, fuelled by either the NHS (yes, I know there is no longer one NHS), or speciality groups (my organ is bigger than your organ….). Third, much of higher education has not caught up with its audience (pace, elite higher education).

Finally, the sorts of experimental science he is talking about (final para) is exactly the sort I love. This is the sort of work Brenner and Crick described when they and a handful of others invented molecular biology. Or for an example from another field, look at how David Hubel and Torsten Wiesel described their work. But sadly, most medical research is no longer like this. It is much, much duller, and much less intellectually  secure, because many built in tests of veracity through experimental design and approach, have been replaced by audit of process. Does it have to be this way? A silent bleak voice, tells me yes. As for the teaching, that is a problem we can do something about.

[I like his use of ergonomic, too]