An important factor in genetic linkage analysis is the penetrance, or the probability of disease onset given the genotype. The impact of variable and age-dependent penetrance on nuclear family linkage analysis and on sib-pair analysis is examined.

In nuclear farnilies, a systematic evaluation of the effects of penetrance misspecification for age-dependent models is reported. Bias in the estimate of disease gene location, and the loss in power associated with poor penetrance estimates are evaluated, and some guidelines are recornmended. When the disease cannot occur in the absence of a particular gene (a necessary gene), it is shown that the estimate of gene location is nearly orthogonal to the parameters of the age-dependent onset functions when the recornbination fraction is small.

In sibling pairs, one methodology available for testing for linkage involves estimating the identical by descent ailele sharing, and calculating a likelihood ratio test for deviations of the sharing from null values. This approach is extended to allow for variability in linkage evidence with covariates. The performance of this extension is evaluated through simulations, and these show that the power to detect linkage can be increased with covariates. In addition, covariate models provide flexibility and greater understanding of the disease process, and aLlow the detection of gene-environment interactionsA second extension of sibling pair methodology ailows information Lom unaffected

siblings to be used when testing for linkage. Because unaffected siblings may develop disease in the future, penetrance corrections must be included which ailow for potential disease suscep ti biLi ty, but the extended mode1 has considerable robustness to misspecification of the susceptibility parameters. Some power improvements are seen in simulations. Empirical methods to estimate the significance levels when testing for Linkage or heterogeneity are also developed.