Nonparametric quantile regression estimation for functional data with responses missing at random (Q2202046)

A functional nonparametric quantile regression model can be used to model the quantile of a scalar response variable with an unknown smooth regression functional of a predictor which takes values in a semi-metric abstract space. This model is more robust than the classical nonparametric functional regression model which models the mean of the scalar response variable. In some practical problems, the response variable may not be observed for some predictors. For the classical nonparametric functional regression model, a nonparametric inverse probability weighting technique was applied by \textit{Q. Wang} and \textit{Z. Sun} [J. Multivariate Anal. 98, No. 7, 1470--1493 (2007; Zbl 1116.62042)] to estimate the regression functional by assuming response was missing at random. In this paper, a similar technique was applied to derive an estimator in a functional nonparametric quantile regression model under the same assumption. Asymptotic normality of the estimator derived is established. The variance in the asymptotic distribution involves a few unknown functions but there is no method proposed to estimate them. Therefore, these results may be difficult to apply in practise. Simulations have been conducted to evaluate the bias and mean square error of the derived estimator, not the difference of the empirical distribution of the estimator and the asymptotic distribution derived by the authors.