st: command xtdpd: different sets of Instrumental variables get different estimated results

[Jian Zhang <jian32@gmail.com>]

Dear Statalisters,

I was using the command xtdpd to estimate the coefficient of
first-lagged dependent variable. The problem I encountered was that
with two different sets of instrumental variables, i got dramatically
different estimates of the coefficient. One estimate is about one
fourth of the other estimate and both estimates seem to be possible
and in reasonable range. Both pass the Sargan test of over-identifying
restrictions.  However, with one set  of IVs, most control variables
are statistically significant while most control variables
insignificant with the other set of IVs. I was wondering how one can
tell which estimate should be preferred. Note that since the two sets
of IVs are overlapped (the number of IVs are in fact same for the two
sets), it is not valid to calculate difference-in-Sargan test
statistics and consider this to be Chi-2 distributed  with the degrees
of freedom equal to the number of the additional instruments.

Any comments/suggestions are really appreciated!

Jian
*
*   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/