Genetic linkage analyses are often performed to identify chromosomal regions harboring genes influencing complex traits. When the trait of interest is quantitative, a method of choice is variance component analysis. This method can be applied to pedigrees of any size and can be easily extended to include environmental factors and multiple genes. However, variance component analysis relies heavily on an assumed normality of the data, an assumption which is often violated when complex traits are concerned. In this talk we present statistical challenges arising from performing genetic linkage analysis in a subset of the Framingham Heart Study subjects consisting of 330 extended families with up to 77 individuals per family. Partial solutions to departure from the normality assumption are presented, including the use of robust score tests and permutation approaches. A particular advantage of the robust score test is that it can be extended to include gene-by-environment interactions. These methods are applied to inflammation and glycemic phenotypes in the Framingham Heart Study.
Group for Research in Decision Analysis