I am using a control function approach with a probit model for the selection equation and a fractional model for the outcome equation.
From the selection equation I calculate the generalized residuals and add them as an independent variable to the outcome equation. For the generalized residuals I use the formula from Wooldridge (2015, p. 428). This formula uses both the estimated parameters and the observed (true) values. Now my question is: how to calculate the generalized residuals when making predictions for a new dataset, for which the true values are not observed?
Wooldridge, J. M. (2015). Control Function Methods in Applied Econometrics. The Journal of Human Resources, 50(2), 420-445.
From the selection equation I calculate the generalized residuals and add them as an independent variable to the outcome equation. For the generalized residuals I use the formula from Wooldridge (2015, p. 428). This formula uses both the estimated parameters and the observed (true) values. Now my question is: how to calculate the generalized residuals when making predictions for a new dataset, for which the true values are not observed?
Wooldridge, J. M. (2015). Control Function Methods in Applied Econometrics. The Journal of Human Resources, 50(2), 420-445.
