Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Panel dynamic OLS


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Panel dynamic OLS
Date   Fri, 6 May 2011 07:37:01 +0100

Good practice on this list is to give full references

http://www.stata.com/support/faqs/res/statalist.html#others

and to explain where user-written programs you refer to come from

http://www.stata.com/support/faqs/res/statalist.html#stata

-xtdolshm- is from SSC.

Nick

On Fri, May 6, 2011 at 7:13 AM,  <hctang@adb.org> wrote:

> I have been trying to find out whether panel DOLS such as that of Kao and
> Chiang (2000) and/or Mark and Sul (2003) be performed in Stata. It turns
> out a third-party contributor has written the Kao and Chiang panel DOLS
> with homogenous covariance structure called xtdolshm.
>
> I have a few general and specific queries which would really appreciate
> some (any) advice:
>
> 1) Can panel DOLS be performed via existing Stata commands like xtreg,
> xi:reg or etc? That is an user just add the lags and leads first
> differences of the suspected cointegrated variables in the estimated
> equation.
>
> 2) Are the standard errors produced in xtdolshm robust to
> heteroskedasticity and autocorrelation? Is this a concern at all in a
> panel time-series estimation as opposed to a typical large N small T
> panels?  If it is and the standard errors are not corrected, can one
> derive (and how) from existing information the Newey hetero and autor or
> any other corrected errors?
>
> 3) If (2) is not feasible or straightforward, would varying the leads and
> lags of the cointegrated variables be a reasonable robustness check on the
> statistical significance of a variable of interest? Say we start with
> 1-lead, 1-lag, then change to 2-lead, 2-lag and so on, to see whether the
> significance of say x1 (in a y x1 x2 x3 equation) changes. We will rely on
> the results of higher leads and lags, presumably because they can better
> address the hetero and autor concerns.
>
> 4) Can the existing xtdolshm incorporate time specific effects or other
> deterministic regressors? If so, how does one write this out?

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index