Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Seliger Florian" <seliger@kof.ethz.ch> |
To | "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu> |
Subject | st: AW: AW: actually vs. reported dropped observations xtlogit |
Date | Thu, 15 Nov 2012 11:10:46 +0000 |
Dear Klaus, It is clear that the number of observations in the fe-model can be lower than with re. My question is why the number of dropped observations reported by STATA (because of no within-group variance in the dependent variable) is higher than the number of observations that have been actually dropped (calculated as N_re - N_fe). Any help is appreciated. Best, Florian -----Ursprüngliche Nachricht----- Von: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Klaus Pforr Gesendet: Donnerstag, 15. November 2012 11:57 An: statalist@hsphsun2.harvard.edu Betreff: st: AW: actually vs. reported dropped observations xtlogit <> Dear Florian, the fe-model drops all observations without within-group variance of the dependent variable, whereas the re-model drops only observation for which the dependent variable can be perfectly predicted with the independet variables. Therefore, the number of observations for fe can be lower as for the re. webuse union, clear xtlogit union age grade, fe xtlogit union age grade, re you can force to use the same estimation sample by restriction to e(sample). webuse union, clear xtlogit union age grade, fe xtlogit union age grade if e(sample), re best Klaus __________________________________ Klaus Pforr GESIS -- Leibniz Institut für Sozialwissenschaft B2,1 Postfach 122155 D - 68072 Mannheim Tel: +49 621 1246 298 Fax: +49 621 1246 100 E-Mail: klaus.pforr@gesis.org __________________________________ -----Ursprüngliche Nachricht----- Von: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Seliger Florian Gesendet: Donnerstag, 15. November 2012 10:50 An: 'statalist@hsphsun2.harvard.edu' Betreff: st: actually vs. reported dropped observations xtlogit Dear Statalist, Using a fixed effects logit model (xtlogit, fe), STATA reports that it drops N=3540 observations because all positive or all negative outcomes. This is understandable given the way how "xtlogit, fe" works. However, in the corresponding random effects model, we have N=4246 observations. In the fixed effects model, we have N=984 observations. 4246-984=3262 ! 3263 is smaller than 3540. Hence, STATA reports to have dropped more observations than it actually did. Do you have an explanation? Thank you for your consideration. Best wishes, Florian * * 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/ * * 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/ * * 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/