Dear Statalists,
I'm estimating a 2-step model, where the first step consists of a FE model (using a long panel) in order to get consistent Fixed Effects (usually called alpha_i hat) in order to regress them in the second step. Particularly, in the first step I take the log of the dependent variable, but in order to give a better sense to my estimates I need to back-transform the results in the original scale. Now, I know there is a vast literature about back-transforming the predicted values (e.g. Manning and Mullahy, 2001), but my problem is that the same kind of reasoning does not apply if I am interested in back-transforming and interpreting the FE, especially after regressing them in the second stage. To be clearer:
1 step) xtreg ln_Y X, fe vce(boot)
predict FE,u
2 step) reg FE Z, vce(boot)
predict P, xb
I need to have P in the original scale of Y.
I would strongly appreciate any help about how to do it in Stata.
Many Thanks
Bellamy
I'm estimating a 2-step model, where the first step consists of a FE model (using a long panel) in order to get consistent Fixed Effects (usually called alpha_i hat) in order to regress them in the second step. Particularly, in the first step I take the log of the dependent variable, but in order to give a better sense to my estimates I need to back-transform the results in the original scale. Now, I know there is a vast literature about back-transforming the predicted values (e.g. Manning and Mullahy, 2001), but my problem is that the same kind of reasoning does not apply if I am interested in back-transforming and interpreting the FE, especially after regressing them in the second stage. To be clearer:
1 step) xtreg ln_Y X, fe vce(boot)
predict FE,u
2 step) reg FE Z, vce(boot)
predict P, xb
I need to have P in the original scale of Y.
I would strongly appreciate any help about how to do it in Stata.
Many Thanks
Bellamy
