Re: st: RE: Areg, absorb

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

<>
On Apr 12, 2011, at 2:33 AM, Jan wrote:

> 
> Would it not be possible for you to first generate group id's based on country1 country2 and year, and then use -areg- to control for each group? I think that would capture the desired effects.
> 
> In that case it would be something like:
> 
> Egen id=group(country1 country2 year)
> Areg ..., abs(id)

That will obviously fail. Imagine a standard panel of N countries and T years. Your group id will have N*T values, exactly equal to the size of the dataset. Adding another dimension of countries doesn't change that. The point of doing two-way or three-way fixed effects is that they independently identify the unit (or units) and the time period, without interactions.

You probably should use this trick to identify country-pairs, though. Do an egen group to create an index of each distinct country-pair, tsset the data by that variable and year, run xtreg, fe with a set of year dummies. Downstream, just use the drop() or keep() options in a program like Ben Jann's estout (on SSC) to retain the coefficients you want to see in the LaTeX table (or whatever).
 
Kit

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




*
*   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/