Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Austin Nichols <austinnichols@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Time to event analysis OLS negative intercept, Survival Analysis |
Date | Tue, 9 Jul 2013 17:41:32 -0400 |
Reese Andorfer <reese.and89@gmail.com>: The log of a number between zero and one is negative, so you are predicting that duration is between 0 and 1 when all X vars are zero... but OLS regression of a log y on X is not appropriate when y can equal zero. See also http://blog.stata.com/2011/08/22/use-poisson-rather-than-regress-tell-a-friend/ and its inspiration http://www.stata.com/meeting/boston10/boston10_nichols.pdf On Tue, Jul 9, 2013 at 5:28 PM, Reese Andorfer <reese.and89@gmail.com> wrote: > Dear all, > > My dependent variable is the duration up until to an event. I did Cox > regressions for my models, which worked fine. Now, since I have uncensored > data I ran OLS regressions ( dependent variable still the duration - > altough I took the log) as a sort of robustness check, however I get now > negative values for the intercepts, which kind of bugs me since I don't > have an explanation for it. Any ideas on what I might be doing wrong ? > > Thank you, > Best Reese * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/