On the Kernel Estimation of the Conditional Mode

The estimator of the conditional mode obtained by maximizing the Nadaraya Watson (NW) kernel estimator of the conditional density function has disadvantages of producing rather large bias and boundary effects. The aim of this study is to overcome these disadvantages by proposing a modified estimator of the conditional mode obtained by maximizing the Reweighed Nadaraya Watson (RNW) kernel estimator of the conditional density function. The asymptotic normality and consistency of the proposed estimator are established and its efficiency is examined by two applications for both simulation and real life data.