Bootstrap Bayesian analysis with applications to gene-environment interaction

We propose a novel statistical model and inferential algorithm for gene environment interaction. Our methodology was motivated by and applied to identity by descent (IBD) sharing for sibling pairs affected by schizophrenia. Our analysis confirms some of the previous findings on the same data set, e.g. the estimated location of the disease gene and the existence of the interaction between the location of disease gene and environment. Our analysis also provides new insights by better accounting for overall variability in the data. We show that taking into account sampling variability may increase the length of posterior credible intervals for the true location of the disease gene by as much as 140%. Moreover, the posterior distribution is shown to be non-Gaussian, which more closely matches the data.