Studies of statistical power to detect linkage: Comparisons of the power of linkage detection for a heterogeneous disease, when using the admixture linkage test vs. the two-locus lod score test, by simulation, and found that even when the two-locus lod score is the correct model for the data, it was generally not more powerful than the admixture test and requires more assumptions about parameters. In a second investigation, it gave no advantage in power of detecting linkage to divide a sample of families into more """"""""homogeneous"""""""" groups by using their posterior probability of being linked to the marker locus. We computed the power of detecting linkage in our sample of 20 bipolar families assuming various levels of heterogeneity. Our sample was very high power to detect linkage if 50% of families are linked and reasonably good power when 25% of families are linked. We have shown that multiple analyses of the same families under different models requires a more stringent lod score criterion. As part of the Genetic Analysis Workshop 7, we have shown that ordering of some closely spaced marker loci is not possible with only one very large pedigree, even when distances between flanking markers are known. We have also shown that for examining linkage of marker loci to malignant melanoma, taking into account familial correlations other than those due to a major locus, can have a large effect on the resulting lod scores.
