Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Ariel Linden, DrPH" <ariel.linden@gmail.com> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | re: Re: Re: st: replicating 2 X 2 data from a paper |
Date | Thu, 4 Apr 2013 10:28:08 -0400 |
Thanks for the follow up, David. I came to the same conclusion that they probably ran -tab- chi2 to get the p values, and they manually calculated the percent difference (no relation to the p value). They certainly should have been clear in how they got to these values, since we're only speculating. Also, to you point about Poisson: it seems that they would have been better served using a count model than contingency tables for no other reason than to provide the reader with contrast estimates with confidence intervals. Testing for over-dispersion is not a difficult task, and it certainly would not take more than a few extra minutes to complete the entire analysis. Interestingly enough, medical journals require that papers reporting results from an RCT show confidence intervals and not just p-values (or more broadly, that authors follow CONSORT statement about writing up results from RCTs). I am not sure how this slipped by. In any case, thanks again for your second pair of eyes on this... Ariel Date: Wed, 3 Apr 2013 20:30:03 -0400 From: David Hoaglin <dchoaglin@gmail.com> Subject: Re: Re: st: replicating 2 X 2 data from a paper Hi, Ariel.