Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Perger, Orsolya" <Orsolya.Perger@afhe.ualberta.ca> |
To | <statalist@hsphsun2.harvard.edu> |
Subject | st: correcting heteroskedasticity in panel data |
Date | Mon, 29 Nov 2010 14:03:50 -0700 |
Hi, I 'd like to get some expert advice on how to correct for heteroskedasticity in panel data. I have a perfectly balanced panel with N=32 group and each of them have T=15 time period. I have to use random effect model , as most of my independent variables are not changing within the paneI. (I am using stata 11, student version. ) I followed the heteroskedasticity test described in the FAQ using xtgls (http://www.stata.com/support/faqs/stat/panel.html ) . The following test I have done: xtgls depvar indvars, igls panels(heteroskedastic) estimates store hetero xtgls depvar indvars local df = e(N_g) - 1 lrtest hetero . , df(`df') result: "Likelihood-ratio test LR chi2(31) = 322.61 (Assumption: . nested in hetero3) Prob > chi2 = 0.0000" So I clearly have a heteroskedasticity problem. (I have tested , no sign of autocorrelation) Based on what I know, I have 2 options: Option1: I can use the white robust standard error to correct heteroskedasticity based on the manual Xtreg depvar indvars, vce(robust) Option2: Or I can use the xtgls, since the LR test supported this version: xtgls depvar indvars, igls panels(heteroskedastic) My problem is that significance of some of my independent variables are different under the 2 options. One of my most important variable(a dummy) is significant under the second option, but not under the first one. I do not know which one is the "true" value. Anybody can offer me some advice? Thanks Orsolya Perger * * 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/