Hello,
I have a question on how to correct the standard
deviations of generated regressors.
If I have a first equation of the form:
1) Z=f(X,Y)
With Z=Exports, X being the exchange rate and Y being
the premium over the exchange rate, if I want to purge
Y from the effect of X, I can run the auxiliary
regression:
2) X=f(Y) and use its saved residuals - u^- in equation
1) to replace Y.
When interpreting the coefficients and t-statistics in
1bis) Z=f(X,u^), I should get the same coefficent for
X, and a different one -presumably lower- on u^.
My question is how do I correct the standard deviation
of u^ whihc might be biased because of inclusion of
any measurement error produced in equation 2)?
Could anyone give me a useful suggestion or reference
on this?
Thanks for your help.
Marcella Vigneri
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|