Kenneth G. Liu <[email protected]> wrote that he obtained different results
between Stata and Limdep for a fixed-effects model with AR(1) errors.
Gao said that ``Limdep uses a (two-stage least squares) 2SLS estimation with
the Cochrane-Orcutt transformation", in which case the estimator underlying
the Limdep command and -xtregar, fe- are different. -xtregar, fe- is using
an ordinary-least squares regression on a Cochrane-Orcutt transformation of
the within-transformed data. Monte Carlo simulations have shown that the
estimator implemented in -xtregar, fe- performs quite well in finite
samples.
If Gao will privately send me his data and log files of exactly what he ran
in each package, we will resolve the issue privately and report back to the
list afterward.
--David
[email protected]
*
* 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/