Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Paulo Regis <pauloregis.ar@googlemail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: two way model and -xtoverid |
Date | Sat, 17 Sep 2011 17:29:34 +0800 |
Dear Statalisteners, I am working with panel data and the distinction between fixed effects, random effects. I am working with -xtreg and -xtoverid (the second command perform a test similar to hausman but valid when you use robust SE). When working with the one-way model there is no problem. I can do tsset ivat tvar xtreg Y X1 X2, robust re xtoverid Then I can compare the result and use the hausman alike test to decide between FE and RE. However, I have problems when I work with the two-way model. If I want to have time there too: xtreg Y X1 X2 i.tvar, robust re xtoverid where "tvar" is a variable with the year of the observation (t=1, ,T). The -xtreg estimate implies FE for years but RE for individuals. The command -xtoverid does not work when I want to test whether I should use FE or RE. If I use dummy variables, I can go around he problem: xtreg Y X1 X2 t2-tT, robust re xtoverid where t2-tT are the dummy variables for T-1 years. However, I raise the issue here to ask if anyone knows if there is any good reason -xtoverid does not want to perform the test. Kind Regards Paulo * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/