Several months ago, the journal Basic and Applied Social Psychology published an editorial announcing a “ban” on p-values and confidence intervals, and treating Bayesian inferential methods with suspicion as well. The editorial generated quite a bit of buzz among scientists and statisticians alike.
In response the American Statistical Association released a letter expressing concern about the prospect of doing science without any inferential statistics at all. It announced that it would assemble a blue-ribbon panel of statisticians to issue recommendations.
That statement has now been completed, and I got my hands on an advance copy. Here it is:
We, the undersigned statisticians, represent the full range of statistical perspectives, Bayesian and frequentist alike. We have come to full agreement on the following points:
1. Regarding guiding principles, we all agree that statistical inference is an essential part of science and should not be dispensed with under any circumstances. Whenever possible you should put one of us on your grant to do it for you.
2. As to specific recommendations on how to do statistical inference, we are in full agreement that you should do it our way. “Our” in this context is understood to represent a subset of us to be worked out at a later date.
Thank you for your consideration.