|
Activity Number:
|
147
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 1, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
WNAR
|
|
Abstract #318142
|
View Presentation
|
|
Title:
|
Estimation and Prediction in the Presence of Spatial Confounding
|
|
Author(s):
|
Garritt L. Page* and Yajun Liu and Zhuoqiong He and Dongchu Sun
|
|
Companies:
|
Brigham Young University and Wells Fargo Bank and University of Missouri and University of Missouri
|
|
Keywords:
|
spatial prediction ;
spatial correlation ;
generalized least squares estimator
|
|
Abstract:
|
In studies that produce data with spatial structure it is common that covariates of interest vary spatially in addition to the error. Because of this, the error and covariate are often correlated. When this occurs it is difficult to distinguish the covariate effect from residual spatial variation (or a spatial random effect). For iid normal errors it is well know that this type correlation produces biased coefficient estimates but predictions remain unbiased. In a spatial setting recent studies have shown that coefficient estimates are also biased, but spatial prediction has not been addressed. We begin a formal study regarding spatial prediction when covariate and error are correlated. This is carried out by investigating properties of the generalized least squares estimator and the best linear unbiased predictor available from a linear model when a spatial random effect and a covariate are jointly modeled. Theoretical results are presented and are accompanied with a numerical study. We demonstrate that in some situations the mean squared prediction error is in fact reduced in the presence of spatial confounding.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2016 program
|
