Many systematic review courses concentrate on interpreting evidence without having to conducting analyses. I am not aware of any "short" courses on this off the top of my head - the ones I am thinking about have been subjects in allied health degrees.

Jim Matthews wrote:

As well as biostatistics issues such as understanding the difference between Risk Ratio and Odds Ratio, a common discussion point with researchers I talk to is when and how to report p values and what do they mean.  Improved education and understanding of this topic would go a long way. 

Aside from noting that the 1 in 19 interpretation for p=0.05 has scant connection to judging the extent to which an alternative of interest is more likely than the null, there seem a plethora of alternatives, with little in the way of common agreement on a substitute.  Myself, in a case where one outcome appears a priori about as likely as the other I say to think of  p=0.05 as corresponding to a relative probability of maybe 2 in favor of the alternative, with a non-point interval null so that one is not testing for the inconsequential.  This issue of the choice of null increases in importance at the sample size increases. 

The R package BayesFactor, which works with a Cauchy prior (t with 1df) and uses a numerical integration approach to avoid MCMC or suchlike sampling, seems to me a reasonable first step on the way a Bayesian approach.  If p lies between 0.005 and 0.05, the Bayes Factors that one gets from BayesFactor seem a reasonable place to go for an indication of what the p-value might mean.  If p<0.005 with a meaninful non-null prior, one is usually safe in thinking that there is a real meaningful difference.

There've been a number of replication/reproducibility studies that provide empirical evidence on what p-values mean in the context of published laboratory studies.  The evidence that they provide should feature strongly in the educational process.

I'm helping create a list of core skills that health and medical researchers need. Two currently on the list are "Introductory statistical analysis" and "Advanced statistical analysis". (I appreciate that these are very broad, but it's a very early stage in the process.)

A suggestion has been made for another skill of "Interpreting statistical analysis" so this is for researchers who rarely run their own analysis, but do read a lot of papers. I'm not sure that this is a good idea, as I've always thought the way to understand outputs is to understand how they were created. But perhaps there is a place for just teaching people about what results mean (e.g., forest plots, t-tests, regression outputs, etc)??