re: st: difference between xtivreg2 and xtreg coefficient estimates

[Christopher Baum <kit.baum@bc.edu>]

<>

Sriniketh said:

I recently ran into an odd issue with xtivreg2. I find that the
coefficient estimates sometimes differ between xtivreg2 and xtreg.
Here are the relevant results from the stata window. You'll notice
that xtreg and xtivreg2 seem to be recognizing a different number of
observations. While xtreg uses 2711 observations, xtivreg2 uses 2341
observations and drops 681 singletons. So the total number of
observations xtivreg2 "sees" is (2341+681=3021). Do you know why this
might be the case?

I should note that this is not always the case. I tried analogous
regressions on a different subset of my sample, and xtivreg2 and xtreg
produce identical results. Moreover, I tried using fixed effects at a
different level for the subset of data used below, and that also
produced identical results between xtivreg2 and xtreg.


It seems very strange that you are getting a different max number of obs/group from the two commands.
Is there a reason you're using the "i" option rather than xtset or tsset?

Why don't you try something like

by coop_sp_qy_ft: drop if _N == 1
by coop_sp_qy_ft: g en = _n
tab en

en should have a min of 2 and a max of ???  (23 or 311 ?!?)

Try running the two estimation commands on these data.

When I try this on a tweaked version of grunfeld, the coeffs are always the same between the two commands, 
but the SEs differ, as xtreg thinks there are nine dummies in the model. But even after dropping the singletons,
the SEs will differ unless you use the 'small' option on xtivreg2 (Schaffer, SSC).

Kit

Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                             An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
  An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   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/