Abstract:

This paper studies the uniform convergence rates of Li and Vuong's (1998) nonparametric deconvolution estimator and its regularized version by Comte and Kappus (2015) for the classical measurement error model, where repeated measurements are available. Our assumptions are weaker than existing results, such as Li and Vuong (1998) which requires bounded support, and a specialization of Bonhomme and Robin (2010) which requires the existence of moment generating functions of certain observables. Moreover, our uniform convergence rates are typically faster than those obtained in these papers.