Thank you for responding. I will try to clarify. The regression
contains time-specific fixed effects only since control variables vary
both on country-specific and country-pair levels. There is one
reporter and multiple partners. Imports and exports are regressed
separately, so presumably i and j indices are "meaningfully
different". It is not possible to define id based on i and j since it
will be constant over categories of goods. The -areg- command does not
seem to handle the estimation correctly since it is not capable of
recognizing how panels are constructed. Any suggestions to move
forward with -xtreg- would be appreciated.
Again, thank you!
Andreas
------------------------------
Date: Thu, 6 Aug 2009 13:45:20 -0400
From: Austin Nichols <[email protected]>
Subject: Re: st: Panel data regression
Andreas Hatzigeorgiou<[email protected]> :
Pooled regression needs no -xtset-; do you want fixed effects?
See -help areg- for an alternative estimator, or -findit felsdvreg- etc.
Are the i and j indices meaningfully different, or are these pairs of
countries where identities are interchangeable?
E.g. if you have exports and imports, the exports of country i to
country j are presumably the imports of country j from country i, so
perhaps you want fixed effects for each pair of countries, not each
country.
You seem to have used i to mean at least two different things in your
email; perhaps you can clarify.
On Thu, Aug 6, 2009 at 1:29 PM, Andreas Hatzigeorgiou<[email protected]> wrote:
> Hi,
>
> Insights on the the following problem would be greatly appreciated.
> The goal is to run a pooled panel data regression using xtreg. The
> data contains reporters (i), partners (j), years (y) and bilateral
> flows (f_i), where i is number of categories of different types of
> goods. There is not a problem running xtreg on the aggregate flow
> level (sum of flows by years). But on the disaggregate level it is not
> possible to use xtsset (partner year) due to "repeated time values
> within panel". I know that identifier and time variables need to
> uniquely identify the data, but in this case I can't see how this is
> possible since the identifier (j) and the time variable (y) will be
> constant over the different categories of goods traded. Is there a way
> around this problem so I still can use xtreg on disaggregate data?
*
* 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/
