[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: FW: Getting F stats for xtivreg
> -----Original Message-----
> From: email@example.com
> [mailto:firstname.lastname@example.org] On Behalf Of
> Freely, Joshua
> Sent: 21 November 2006 21:55
> To: email@example.com
> Subject: st: FW: Getting F stats for xtivreg
> Hello StataList,
> I am running an xtivreg, re and I am trying to measure the
> validity of my instruments. Does anyone know of a way for me
> to obtain an F-statistic for my first stage equation? I
> know xtivreg2 does it for FE but not RE models. Anyone have
> any ideas for me?
I'm working on a new version of xtivreg2 that will do random effects models and other things as well, but in the meantime there's an alternative, namely the new command -xtoverid- (available from ssc-ideas in the usual way).
The way -xtoverid- works internally is by reestimating, via a call to -ivreg2-, whatever -xtivreg- or -xthausman- equation was just estimated.
The -xtoverid- command has an undocumented -noisily- (or just -noi-) option. This tells -xtoverid- to display the internal -ivreg2- reestimation results. For the purposes of the -noi- option, the call to -ivreg2- includes the -ffirst- option, so you will see the usual array of first-stage/weak/underidentification stats (but not the first stage regressions themselves).
There is a small problem, which is that the internal reestimation uses temporary variable names. Thus the first-stage stats will be for variables with names like __00000P. These map 1-for-1 to the original variables, however, and there's a good chance you should be able to work out what they are.
So... after your -xtivreg- estimation,
should tell you what you want to know.
Hope this helps.
> Thank you.
> Joshua L. Freely
> Research Associate
> Public/Private Ventures
> 2000 Market Street, Ste. 600
> Philadelphia, PA 19103
> 215.557.4425 (office)
> 215.557.4469 (fax)
* For searches and help try: