Abstract
A method of obtaining weak or strong invariance principles for the centered and normalized deconvoluting kernel estimator is presented. It is shown that these approximations, under certain smoothness constraints, achieve the best global rates of convergence.