Abstract:
|
In matched case-crossover studies, the conditional logistic regression model is not able to detect any effects associated with the matching covariates by stratum. However, some matching covariates such as time and spatial location often play an important role as an effect modification which makes incorrect statistical estimation and prediction. Hence in this paper, we propose two semiparametric spatio-temporal varying coefficient models to evaluate effect modification by time and spatial location in order to make correct statistical inference. Our proposed models are developed under the Bayesian hierarchical model framework.
We briefly introduce a method which allows us to simultaneously evaluate parametric relationships between the predictor and binary outcomes, semiparametric relationships between the predictor and time, and effect modification due to spatial locations for an appropriate number of spatial locations. We present our second approach in detail which allow us to evaluate the same relationships, but in the presence of a small number of locations. We illustrate the application of this approach with a 1-4 matched case crossover study of aseptic meningitis in children.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.