Revisiting Gaussian copulas to handle endogenous regressors