Abstract: This talk intends to show empirical examples that deal with the problem of controlling unobserved heterogeneity in nonlinear models. In some applications unobserved heterogeneity is dependent of the explanatory variables making much more difficult to find a consistent estimator of the unknown parameters. In this context two different approaches are followed. One, uses models with random coefficients and the other semiparametric methods. Applications with cross-sectional and panel data are presented.