The American Statistical Association released a statement yesterday on the (mis)use of p-values. It includes 6 principles on p-values which I think are worth repeating and spreading:
- P-values can indicate how incompatible the data are with a specified statistical model.
-
P-values do not measure the probability that the studied hypothesis is true, or theprobability that the data were produced by random chance alone.
-
Scientific conclusions and business or policy decisions should not be based only onwhether a p-value passes a specific threshold.
- Proper inference requires full reporting and transparency.
-
A p-value, or statistical significance, does not measure the size of an effect or theimportance of a result.
-
By itself, a p-value does not provide a good measure of evidence regarding a model orhypothesis.
Like Tim Haab (see his post and take on p-values, to which I agree, here) I am not much of a statistician or econometrician. I gladly leave that to others in our research group. However, I do teach a course in advanced quantitative methods in our M.Sc. program in cultural sociology, and that requires of course that I explain the p-value. If students of that class read this blog, I hope they recognize the above principles at least partly from my classes, even though I have not treated the principles separately and explicitly.
Overall, I do tend to try and make clear that:
- Statistical significance is not sociological (or economic, or biological, or …) significance
- The absence of proof is not a proof of absence.
- Correlation is not causation.
- The importance of full reporting and transparency.
Incidentally, the last principle is one of the main reasons why the course is taught in R; it makes the results more readily replicable, although it generates an additional complication:students have to learn how to program, in addition to learning about p-values.
*Yes, I know it’s a terrible pun, but I do hope that it helps boosting the effect size of the ASA-statement
Illustration credit: Tyler Vigen. http://tylervigen.com/spurious-correlations