The journal Basic and Applied Social Psychology is banning the NHSTP. (That’s the “null hypothesis statistical testing procedure” you might remember from an intro stats course.) This includes banning confidence intervals, thanks to the duality between confidence intervals and hypothesis tests. The journal’s editors write that:
We hope and anticipate that banning the NHSTP will have the effect of increasing the quality of submitted manuscripts by liberating authors from the stultified structure of NHSTP thinking thereby eliminating an important obstacle to creative thinking.
I’ve seen the fixation on statistical significance be a big block in presenting results in business settings. I can’t count the times where I’ve had to explain, especially with “big data” sets, that just because something is statistically significant doesn’t mean it’s practically significant. How much of the cult of statistical significance comes from the choice of words? Presumably the same is true in the social science setting; at least in both cases you have people with some statistical education and who are generally used to looking at numbers but are not statisticians or otherwise specialists in a quantitative field.
However this may be a bit of an overreaction. To say “thou shalt not do X” could be just as restrictive as “thou shalt always do X”.
3 thoughts on “Psych journal bans hypothesis testing”
I think it a huge over-reaction. It seems like they realize that p-values are being abused (e.g. hunting for p<0.05) and so their reaction was to ban them. A better reaction would be to simply try and better understand p-values, imho. Cheers!
It seems like an overreaction to me, too, although I don’t know how badly p-values are being used in the literature and sometimes a debate about publication standards needs to be inspired by a big and apparently over broad move.