Statistical methodology for testing genetic association in family-based studies

Sammanfattning: This thesis is concerned with family-based studies of association between genetic markers and binary traits. A special point of interest in family-based association studies is to separate the within family correlation and the genetic effect common among all families in the study. In family-based studies of quantitative traits within family trait correlation is explicitly modeled in terms of both alleles shared identical by descent and common environment. We have extended this notion to a binary trait setting, and formulate a generalized linear mixed model based on a log-log link and gamma distributed random effects capturing the within family correlation induced by linkage. The genetic effect common among all families is captured in the linear predictor of the model. We show that the model can be used to construct tests for a variety of situations; for testing association between single markers and a trait, for testing association between multiple markers (jointly) and a trait, and for testing association between a single marker and two diseases jointly. We have evaluated the model in four papers and show that the power of the test is up to double that of the gold standard for testing association in the presence of linkage - the Family-Based Association Test (FBAT).