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

st: WLS


From   "Antonio Rodrigues Andres" <ara@sam.sdu.dk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: WLS
Date   Thu, 05 Aug 2004 18:47:38 +0200

Dear Stata users
Assume that you have have as dependent variable
H, the natural log of the mortality rate. Cook and Tauchen (1984)
demonstrate that this model has heteroskedasticity which takes the form
ln (1+(1-p)/pN) where p is the probability of death and N the cell size.
Thus, weigthing by (the square root of) population size is not correct.
However, with small p, the variance of the error term is approximately
proportional to the cell size, and the procedure is close to being
right. In this case 1-p approx = 1 and for large N, ln (1+(1-p)/pN)
approx 1/pN

This means that we should use analytical weights like
regress H x1 x2 [w=pop] is that correct?

I would appreciate your suggestion
Sorry for any inconvenience
Regards
Antonio
*
*   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