Consistent estimation of limited dependent variable models despite misspecification of distribution (Q580866)

This paper focuses on estimation when observation of the true dependent variable, y, is imperfect. Another class of problems we will examine are characterized by observation of y up to an unknown, but monotonic increasing, transformation. The popular methods of estimation for limited dependent variables models (LDV) specify the error distribution and their robustness against misspecification is an open question. In this paper, the author proposes a weighted M-estimator (WME) for these estimation problems. This estimator is consistent, up to an unknown scaling factor, for the slope parameters in the regression function on the imperfectly observed dependent variable. The WME is attractive because it has weak requirements for knowledge of the relationship between the observed dependent variable and its latent counterpart and because the computation is a modification of popular estimators.