Finite sample performance of deconvolving density estimators

摘要: Recent studies have shown that the asymptotic performance of nonparametric curve estimators in presence measurement error will often be very much inferior to when observations are error-free. For example, deconvolution Gaussian worsens usual algebraic convergence rates kernel slow logarithmic rates. However, mean large sample sizes may required for asymptotics take effect, so finite properties estimator not well described by asymptotics. In this article calculations performed important cases and Laplacian which provide insight into feasibility deconvolving density practical sizes. Our results indicate lower levels can perform reasonable