Workshop on Structure Adapting Methods - Abstract
Mammen, Enno
We discuss nonparametric regression models where the covariates are only partially observed. We assume that a nonparametric estimate of the covariates is available. This is a standard situation in recent nonparametric models in econometrics, see e.g. simultaneous nonparametric equation models with or without additive error, estimation of treatment effects by regressing on fitted prospensity scores, nonparametric GARCH-in-means models, iterative nonparametric estimation schemes. We present general results for two step estimators based on regressing on the estimated covariates and discuss the effect of the first step on the asymptotic distribution of the resulting estimator.