tag:blogger.com,1999:blog-4900303239154048192.post7946863854901714692..comments2024-03-06T06:34:42.881-05:00Comments on EconoSpeak: Statistical Significance and the Sweet Siren of Self-Confirmation: A Reply to TaylorUnknownnoreply@blogger.comBlogger3125tag:blogger.com,1999:blog-4900303239154048192.post-47328774688733470862019-04-22T20:08:31.169-04:002019-04-22T20:08:31.169-04:00ps: As a self-disclosure, I'm one of the signe...ps: As a self-disclosure, I'm one of the signers of the <i>Nature</i> letter.Peter Dormanhttps://www.blogger.com/profile/00093399591393648071noreply@blogger.comtag:blogger.com,1999:blog-4900303239154048192.post-12761353182420431452019-04-22T20:06:59.463-04:002019-04-22T20:06:59.463-04:00Hi Bark. Actually, I didn't mention the issue...Hi Bark. Actually, I didn't mention the issue of effect size vs p-value at all. I largely agree with them, but the focus of this post, and the larger current debate over significance testing, is about what constitutes evidence. I am aware of the proposal to cut the cutoff down to .005, but that strikes me as utterly wrong-headed. It doubles down on the power of the lexicographic filter, as if the only problem is that we haven't been putting enough weight on that initial significance test. No, get rid of the filter and judge the weight of evidence substantively, putting p-values in the hopper along with the other factors. Do read the <i>Nature</i> piece.Peter Dormanhttps://www.blogger.com/profile/00093399591393648071noreply@blogger.comtag:blogger.com,1999:blog-4900303239154048192.post-16886046473599039532019-04-22T19:54:53.735-04:002019-04-22T19:54:53.735-04:00Peter,
You briefly sccarfed over what McCloskey a...Peter,<br /><br />You briefly sccarfed over what McCloskey and Ziliak have said about this, your "effect size." The strength of the relationship is more important than the p-value.<br /><br />It must be noted that while some are joining the M-Z critique of stat significance, there are some outside of economics who are effectively doubling down and criticizing Fisher's focus on 5% because it is too large a number, that the asterisks should only be dragged out for much stronger levels of significance. However, most of those people are in hard sciences where data is not nearly as noisy as it is in economics and other social sciences.rosserjb@jmu.eduhttps://www.blogger.com/profile/09300046915843554101noreply@blogger.com