Speaker：Prof. Chun-Shu Chen (Institute of Statistics, NCU)
Topic：An advanced approach to unmeasured spatial confounders
Speaker：Prof. Chun-Shu Chen (Institute of Statistics, NCU)
Date Time：FRI. Apr 30,2021, 10:40 AM - 11:30 AM
Place： 4F-427, Assembly Building I
Abstract
In spatial regression analysis, collinearity between covariates and spatial random effects can lead to significant bias in the fixed effect estimate. This issue is known as spatial confounding. A reliable inference is difficult due to unobserved spatial random effects. Some techniques under restricted spatial regression had been proposed to investigate this issue, but how to modify the bias of regression coefficient estimators remains an active research topic. We propose an adjusted generalized least squares estimation method to estimate regression coefficients in the presence of spatial confounding and its superiorities are demonstrated by theories and simulations. Some concerns about the proposed approach are also discussed. Finally, a real data example is analyzed for illustration.