Comments for 0 at http://www.cepr.net , comment 1 to 1 out of 1 comments
http://www.cepr.net
Mon, 20 Oct 2014 17:55:57 +0100FeedCreator 1.7.3You have writtens something false
http://www.cepr.net/index.php/?catid=#comment-9014
"Our standard of significance is a result that you would not get by random chance more than 10 times in a hundred. This means that if you ran 1000 regressions of things that had nothing to do with each other, you would expect 100 of them to have statistically significant results."
Statistical significance is about re-sampling, not re-specification.
If you have a seven potential control variables (which I think gets you to about the 1000 possible regressions) but they are all highly collinear, the chances are very high you will get the same results for all of them. Ditto if you run OLS, then a Tobit, etc. The probability of a regression being robust in this way has nothing to do with sampling distributions, which is what that damnably ubiquitous "statistical significance" is all about.
- Curmudgeonly pedantThu, 19 May 2011 00:11:02 +0100