On classification with nonignorable missing data (Q2034467)

\textit{J. K. Kim} and \textit{C. L. Yu} [J. Am. Stat. Assoc. 106, No. 493, 157--165 (2011; Zbl 1396.62032)] proposed an efficient tool to estimate unknown population parameters based on nonignorable missing data with a semiparametric exponential tilting propensity. Such missing response mechanism was employed in this paper to construct new kernel classification rules that can be used to perform statistical classification with nonignorable missing responses. The strong asymptotic optimality of the proposed kernel classifiers is established.