Robust transmission regression models for linkage and association

We present a general regression model that accounts for both linkage and linkage disequilibrium (LD) when analyzing nuclear family data. The method does not require LD to exist in order to evaluate linkage, but if LD does exist, the power to detect linkage can increase due to improved information on linkage phase. The proposed method is general, allowing for a variety of traits (e.g., binary affection status, categorical and quantitative phenotypes), affecteds only analyses, and covariates. Covariates can be useful to assess heterogeneity of linkage and LD, as well as gene-environment interactions. Other advantages of our methods are that: LD parameters are not defined without linkage, so that population stratification cannot bias the analyses; a combined test for linkage and LD can be used to test for linkage; given the existence of linkage, an adjusted LD test useful for fine-mapping can be constructed; covariate effects can be flexibly modeled; and families containing a single child and families containing multiple offspring can be combined for a single analysis (capitalizing on the LD information provided by single-child families and the combined linkage and LD information provided by multiple offspring). The basic features of the regression model are presented, as well as discussions of potential applications and critical statistical issues.