st: Correcting for heterogeneous measurement errors in linearregressions

[Laura Gibbons <gibbonsl@u.washington.edu>]

In searching the archives I found my exact question, but no answer.  I've 
not heard back from the author, so thought I'd try the list again.

I've got ability estimates from item response theory, so that I know each 
person's estimate and the error in that estimate, and the error changes 
from person to person.  For more details, please see below.  I'd like to 
be able to use that error term directly in a regression model (rather than 
some sort of bootstrap).  I appreciate any help or direction anyone can 
give.

many thanks- Laura
--------------------------------------------------------------------------------

Dear Statalisters,

I'd like to run a regression:

yi = a + � * xi + ei

where xi is a true, but unobserved, independent variable.

The observed explanatory variable, which I estimated in a previous
regression, is measured with error:

Xi = xi + ui.

where Xi is the estimated regression coefficient and ui is the measurement
error in Xi with mean 0 and a known variance ?2u,i (this variance is equal
to the square of the estimated standard errors reported on the estimates
Xi).

I know that there are two STATA commands that deal with measurement error,
EIVREG and RCAL.  However, I think that both commands allow only for 
constant measurement error, whereas my measurement error is heterogeneous 
and varies with observations.  How can I correct for heterogeneous 
measurement errors in STATA?

Thank you very much for your help!

Best regards,
Jan






~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Laura E. Gibbons, PhD
General Internal Medicine, University of Washington
Box 359780
Harborview Medical Center
325 Ninth Avenue, Seattle, WA  98104

phone: 206-744-1842   fax: 206-744-9917
Office address: 401 Broadway, Suite 5122
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~