re: Re: st: AW: Autocorrelation after xtdpdsys

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

<>
Angela said, re DPD estimation,

It still bothers me though, what if we have autocollelation of second
or higher order?

Is there a way (a command or routine) to correct it and take new
autocorrelation-corrected coefficients?

Autocorrelation of second or higher order in the residuals may signal that some of the orthogonality (moment) conditions upon
which you are relying for consistency of the estimates are not being satisfied. For instance, with AR(1) errors, second lags of an endogenous variable are not valid instruments (while third lags are). If you use second lags, you will probably reject the null
of the AR(2) test.

An excellent source of detailed information on these topics, even if you're using the official Stata commands, is David Roodman's 
"How to do xtabond2" article in the Stata Journal (2009), which given its age is freely downloadable. This paper and another, on -cmp-,
recently won the Stata Journal's first Editors' Prize.


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
                                                                                                   | http://www.crup.com.cn/Item/111779.aspx	


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/