I’m been meaning to write something non-trivial, but until then here’s something to entertain you. I’ve not had experiences this bad, and most of my collaborators are really good about discussing things, but sometimes…
I think the problems come when the statistician isn’t seen as a collaborator, but as a workman or a servant. The Beast thinks of me as a servant too, but he never annoys me with questions about why he can’t just do a t-test.
Oh, and I wish my office was as spacious as the one depicted.
(h/t Orac)
I am not a statistician, but I still feel your pain. The number of biology researchers (including clinical researchers) who understand the whys, wheres and whats of biostatistics is infinitesimally small. Douglas Altman has lamented about this. There was an article in one of the American Society for Microbiology journals, that called for better statistical analysis to be done and presented in papers; it condemned the rather common and erroneous practice of depicting means and SD/SEM on bar graphs in figures, and analyzing that data with a non-parametric test which does not look at means, such as the Mann Whitney U test. Sadly, this practice still exists, even in journals as hallowed as Nature.
Much worse is the heavy emphasis on p-values in problems that are really about estimating parameters (which require confidence intervals, not p-values, for tehir proper interpretation). I hope some day the statisticians of the world rise up and revolt, and refuse to do another silly p-value on an artificial null hypothesis that we could reject without leaving the office. Statisticians, science is in your hands! Rise up!!
Oh gods, I’m with you on that, Lou. The problem is that the poor biologists would have to think instead.
Back when I started in research in the mid-80s it was relatively commonplace in much of biology to publish papers with no statistics at all…. happy days. For instance, my first published paper from 1987 is a joyously stats-free zone.