Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: xtivreg, fe , multiple reg with interaction effects


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: xtivreg, fe , multiple reg with interaction effects
Date   Tue, 2 Nov 2004 13:15:28 -0500

> 1) I want to perfrom multiple linear regression of the
> whole model with categorical predictors, i.e. look at
> the interaction effect between time/region dummies and
> ALL continuous other variables. Are there any sources
> in Internet to refer to?

This does not seem like a very sensible thing to me: if you have 5
time points, 100 regions, and 8 explanatory variables, then you would
need to create 800 interactions, which is far more than 500
obserations that you have. Could you please clarify what it is exactly
that you want to do.

> 2) When I use xtivreg, fe command to do 2SLS with
> instrumental variables, significance of the
> coefficients is shown with a z-statistic, just like in
> the random effects case, although I expected to see
> t-statistics as in xtreg, fe. Is there anything wrong
> I could be doing?

The inference based on instrumental variables estimators is only valid
asymptotically, since the only nice statistical property that IV
estimators have is consistency. Hence the z- rather than t-statistic.
You can play around with your degrees of freedom to inflate it
somewhat, and get more accurate estimates. I think this issue has been
arising from time to time at Statalist, and it is mentioned briefly in
xtivreg FAQ you mentioned below.

> 3) These significance levels of the coefficients come
> out very very small (almost die out) in xtivreg, fe
> case. Could it be due to any problem I am not aware
> of?

Not quite sure I am following you. Give a piece of output to compare.

> 4) I found in STATA FAQ the following question "How
> can I estimate a fixed-effects regression with
> instrumental variables?" and answers refer to ivreg
> command through indicator variables. What is the
> difference between the case in this question and the
> xtivreg, fe? is it only about STATA6 version that did
> not have xtivreg, fe?

I think this is a version question: the FAQ was written in 2001, when
Stata 6 only started shipping.

-- 
Stas Kolenikov
http://stas.kolenikov.name
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index