Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | lucie gadenne <gadenne@pse.ens.fr> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: gmm problem with e(sample) and xtinstruments |
Date | Fri, 12 Nov 2010 19:24:42 +0100 |
I'm new to using gmm with panel style instrument, and I have the following problem. When I run the regression : "gmm (y - {xb:x1 x2}- {bo}), instruments( x2) xtinstruments(z, lags(0,2)) winitial (xt L)" I get standard estimation results, except that I am told my number of observations is N, where N=number of observations in my dataset. This shouldn't be as some of my observations (roughly 10%) have by construction no value for the instrument z, whatever the period. My understanding is that the gmm command should drop the observations for which none of the xtinstruments are available, and I do want to drop them. I then type : "count if e(sample)" I get 0 observations. If however I type: "count if !e(sample)" I have all my observations. And finally "dis e(N)" gives me N, the total number of observations. So I have two questions: _why doesn't gmm exclude the observations for which none of the lags of the instrument is available? _why is e(sample) not marking any observations as part of my estimation sample? Thanks a lot for your help, Lucie -- Lucie Gadenne PhD Student Paris School of Economics * * 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/