A kernel-type regression estimator for NMAR response variables with applications to classification

This work deals with the problem of nonparametric estimation of a regression function when the response variable may be missing according to a not-missing-at-random (NMAR) setup. To assess the theoretical performance of our estimators, we study their strong convergence properties in L p norms where we also look into their rates of convergence. We also study applications of our results to the problem of statistical classification in semi-supervised learning.