|
Abstract:
|
We describe a novel application of stochastic search variable selection to gene mapping of complex diseases (Swartz et al, Biometrics, 2006). This method introduces a dual hierarchy indicator prior that mirrors selecting alleles at loci. We also exploit the inherent hierarchical structure of SSVS applied to conditional logistic regression to incorporate linkage disequilibrium, a form of genetic correlation, into the model. We show that incorporating both the dual hierarchy indicators and the genetic dependence structure reduces the false positives and false negatives over conditional logistic regression alone. We show the methods performance on real and simulated data. Our software is available at www.epigenetic.org/Linkage/ssgs-public.
|
