Estimating the Distribution Function of a Stationary Process Involving Measurement Errors

A nonparametric estimation of a distribution function is considered when observations contain measurement errors. A method is developed to establish asymptotic normality results for a deconvoluting kernel-type estimator for ρ-mixing stochastic processes corrupted by some noise process. It is shown that the asymptotic distribution depends on the smoothness of the noise distributions, which are characterized as either ordinary smooth or super smooth. Also, the kind of dependence of the noise process is crucial to the form of the asymptotic variance.