Title: Nonparametric estimation in spatial regression
Authors: Tatiyana Apanasovich - George Washington University (United States) [presenting]
Abstract: It is well known that the selection of the smoothing parameter in nonparametric regression is difficult when the errors are spatially correlated. We will discuss various smoothing parameter selection procedures which require a prior knowledge about the correlation structure. Next, we propose a regression estimation framework based on modified kernels which is less sensitive to correlated errors and requires little to no prior knowledge about their correlation structure and its parameters. We demonstrate the practical value of the proposed methodology through simulation studies and real data analysis.