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

Re: st: xtabond problem

From   "sistoand80" <[email protected]>
To   "statalist" <[email protected]>
Subject   Re: st: xtabond problem
Date   Fri, 1 Jul 2005 18:16:42 +0200

Dear Rashmi, 
have you tried using xtabond2 (ssc install xtabond2)? xtabond2 is more feasible in the instruments setting, allowing a drastical cut in the number of instruments through the options "gmm" and "iv". For example, if you want to instrument yor lagged dependent variable with only one instrument for cross, you can type "gmm(l.depvar, lag(2 2))" and your endogenous variable will be instrumented with only the t-2 observations. If you use also lags of predetermined variable as instruments, you can use the same proceedment (gmm option) or treat them as esogenous variable using the option "iv".

Andrea Sisto
Phd student
University of Eastern Piedmont (Italy)

> Rashmi Shankar <[email protected]> has a panel dataset with many
> observations per panel.  When Rashmi runs -xtabond- on this dataset, the
> estimation takes a long time.
> The Arellano-Bond estimator was designed for datasets with a small number of
> observations per panel.  For datasets with many observations per panel, one
> can use -xtivreg-.
> As noted at the bottom of page 37 of the XT manual, the number of
> instruments used in the Arellano-Bond estimator grows rapidly with the
> number of observations per panel, causing the estimation to take a long
> time.
> Rashmi could use the -maxldep()- to limit the number of lags for which
> instruments are created.  However, datasets with lots of observations per
> panel have too many instruments, even when setting -maxldep(1)-.
> If Rahsmi has lots of observations per panel, I suggest that Rashmi look at
> [XT] xtivreg for an example of another estimator for the same parameters.
>      --David
>        [email protected]
> *
> *   For searches and help try:
> *
> *
> *

6X velocizzare la tua navigazione a 56k? 6X Web Accelerator di Libero!

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index