|
Abstract:
|
When performing haplotype-based analyses in case-control studies, appropriate inference is based on a retrospective likelihood. However, prospective analysis is typically easier as it does not require an explicit model of the joint distribution of haplotype and environmental factors. As such, it has become common practice for data to be analyzed prospectively even though it was collected retrospectively. Although prospective methods have been reported to be valid but less efficient than their retrospective counterparts when testing for haplotype-phenotype association, ignoring the sampling scheme can result in substantial bias when estimating the haplotype effects. In this work, we examine the consequences of ignoring the sampling scheme when estimating haplotype effects via penalized regression, which combines the estimation and testing of genetic effects, in case-control studies.
|
