Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: two stage model variance estimators


From   Michael Hanson <mshanson@mac.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: two stage model variance estimators
Date   Sat, 17 Sep 2005 08:11:46 -0400

On Sep 16, 2005, at 8:04 AM, Rachel Bouvier wrote:

The original project was to try and get a "trend" of GDP over
time, with the residuals as the fluctuations around that trend. Then, I
would use both the trend and the flucuations as regressors in my second
stage model. Again, this was a few years ago, and I am trying to dust
off the model and see if it makes sense to revisit it. I know a lot
more now than I did then - I'm going to have to rethink the whole thing.
I wish I had consulted Statalist then!
If you have enough data in the time dimension to estimate separate trends per country, you may want to consider taking natural logs of your GDP data and then linearly detrending (i.e. regressing against time instead of both time and time^2). Odds are GDP exhibits compound growth, so using a quadratic is just an approximation anyway -- and not necessarily a good one -- to an underlying exponential growth function. Additionally, when you use logs your estimated coefficients will have elasticity interpretations, which are useful (as they are unit-free) in cross-country studies. Hope this helps.

-- Mike

*
* 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/




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