On Thu, 15 Oct 2009, Britt, Michael went:
...then you have to deal with the Results section. Forget about the relatively straightforward t-test or anova. Today's articles are filled with lengthy explanations of the most detailed statistical procedures imaginable (Auto-Regressive Integrated Moving Average anyone?). I taught stats for many years and took a multivariate stats class in grad school and I can only begin to understand the typical results section of some of these articles. And we expect "people-people" to connect with (or even want to read) a typical scientific article in our field?
That's a good point--the kinds of studies that are of greatest interest to clinicians are exactly the kinds of studies that are best analyzed with complex statistical methods, making the reports daunting to read. But this problem can be overcome by good writing. Readers might not understand the mathematics behind a particular technique, but they can understand the *logic* if it's explained straightforwardly. I'll link to one of many examples--in an APA journal: <http://www.ncbi.nlm.nih.gov/pubmed/19254107> This article relies on some heady multilevel modeling, described in detail for those who want detail, but it also contains plain-English explanations of what the models mean. I've shown it to a couple of smart people whose background in stats in minimal, and they found the results both understandable and fascinating. Moral: use whatever abstruse stats you need, but write to be understood! I'd be interested in seeing other TIPsters' examples of papers that pull this off. --David Epstein da...@neverdave.com --- To make changes to your subscription contact: Bill Southerly (bsouthe...@frostburg.edu)
