Two-stage spatial regression models for spatial confounding
Abstract Number:
3630
Submission Type:
Contributed Abstract
Contributed Abstract Type:
Poster
Participants:
Nathaniel Wiecha (1), Jane Hoppin (1), Brian Reich (1)
Institutions:
(1) North Carolina State University, Raleigh, NC
Co-Author(s):
First Author:
Presenting Author:
Abstract Text:
Public health data are often spatially dependent, but standard spatial regression methods can suffer from bias and invalid inference when the independent variable is collinear with spatially-correlated residuals of the response variable. This could occur if, for example, there is an unmeasured environmental contaminant. Two-stage spatial regression models counteract this dependence by removing a nonparametric spatial trend from both the explanatory and response variables before estimating parameters of interest. There has been limited theoretical examination of two-stage regression models applied to spatial data over continuous spatial domains. We review existing theory of semiparametric two-stage regression models, and examine how using Gaussian Process regression to model the spatial trends affects regularity conditions needed for root-n consistency, asymptotic normality, and variance estimation. We compare finite-sample properties of several estimators in a simulation study and find that certain estimators are able to correct for bias and obtain nominal coverage in confounding scenarios where standard spatial regression estimators are highly biased and have poor coverage.
Keywords:
Bias reduction|Semiparametric regression|Spatial confounding|Gaussian Process| |
Sponsors:
Section on Statistics in Epidemiology
Tracks:
Statistical Issues in Environmental Epidemiology
Can this be considered for alternate subtype?
Yes
Are you interested in volunteering to serve as a session chair?
No
I have read and understand that JSM participants must abide by the Participant Guidelines.
Yes
I understand that JSM participants must register and pay the appropriate registration fee by June 1, 2024. The registration fee is non-refundable.
I understand
You have unsaved changes.