I would like to run disattenuated regression (DR) in Stata. I have not
found a package that will do this for me.
So I think I have to manually alter the variance-covariance matrix and
run the regression on that altered matrix. I have never done anything
like this before so any guidance would be really appreciated. I am
running Stata 12 on Windows.
I found the following description of a procedure for manually altering
the variance-covariance matrix (V):
http://www.stata.com/support/faqs/statistics/instrumental-variables-regression/In my case, I need to only correct the variances (for measurement
error), the covariances will not change.
However, in my case the coefficients should change as a consequences of
this correction, while in the example they do not. Does this mean I have
to manually calculate the coefficients and update (b) too? Or is there a
command I can use to estimate new coefficients based on the corrected V?
Can I use an approach similar to what is shown in the example? If not,
what approach can I use?
