Bootstrap based model checks with missing binary response data

Dikta, Kvesic, and Schmidt proposed a model-based resampling scheme to approximate critical values of tests for model checking involving binary response data. Their approach is inapplicable when the binary response variable is not always observed, however. We propose a missingness adjusted marked empirical process under the framework that the missing binary responses are missing at random. We introduce a resampling scheme for the bootstrap and prove its asymptotic validity. We present some numerical comparisons and illustrate our methodology using a real data set.