| In genetic linkage analysis, association between the sharing of genetic material Identity-By-Descent (IBD) and the sharing of phenotypes in related individuals is examined. With incomplete haplotype or genotype information, IBD sharing may not be directly observed. In these cases, observed data are combined with estimates of population frequencies to make inferences about IBD status.; When the estimates of population frequencies are imprecise, or, a fortiori, in the presence of population heterogeneity (where frequencies vary among subpopulations), biased inferences about IBD can result. These biased inferences, in turn, can lead to spurious evidence for linkage. Although methods for adjusting for confounding due to population admixture have been developed in the context of association analyses and for single point linkage analyses, these methods have not been extended to multipoint linkage analyses.; In this dissertation a general approach to adjustment for population admixture is adapted to multipoint linkage analyses. A fundamental issue in extending the general approach beyond single point analyses involves incorporating recombination fractions between markers.; The general approach to adjustment that is adapted here involves linear regression calculations which themselves involve the conditional probabilities of observed genotype data of pedigree members given possible values of the haplotypes of pedigree founders. In some cases, for example, when incomplete genotype data or unknown phase preclude direct observation of the conditional probabilities, the number of predictors in the regressions may be quite large, and substantial computational complexity may result. One result of this dissertation is an approach to reducing the computational burden. The results of simulation studies that illustrate the operating characteristics of the approach are presented. The results reveal that in the presence of population admixture, the approach can avoid spurious evidence for linkage while preserving power. |
