Kernel deconvolution density estimation
Daniel J. Nordman
This dissertation is about kernel deconvolution density estimation (KDDE), which is nonparametric density estimation based on a sample contaminated with measurement error. It is separated in four parts. First we explore some methodological aspects of KDDE. In the following two parts we describe the computational challenges in KDDE and our statistical software for KDDE in R. Finally, we propose a simple bandwidth selection procedure that has good theoretical properties.