Statistics Seminar - 03/28/24

Abstract: In this talk I will attempt to show a couple of important methodological contributions to the ongoing research on probability density function estimation. I will consider semiparametric density estimation setting, defined as the product of an initial parametric approximation and a correction factor.  There the first contribution is that the proposed modification of the correction factor improves estimator’s accuracy in terms expected L-2 error. As a by-product, it also  adds  a couple additional desirable  features into the estimator.   The second is that this work develops mean integrated square error optimal bandwidth selector which depends on the initial parametric approximation and establishes its statistical properties. The improvement in estimation precision will be illustrated via distributional data comparisons. Finally, a real data analysis will be presented to show how the present research helps uncover the true underlying data pattern in practice.