Inference in nonparametric current status models with covariates

Abstract

In interval censored models with current status observations, the variables are indicators of the presence of individuals on observation intervals and covariates. When several individuals share the same observation interval, a simple procedure provides new estimators for the distribution of the observation times and their intensity, in a closed form. They are $n^{1/2}$-consistent for piece-wise constant covariates. Estimators of the sample-sizes are deduced and asymptotic $\chi^2$ tests for independence of the observations on consecutive intervals and for independence between consecutive classes for the observed individuals are proposed.