Because our analyses were blind to the generating model, we began our investigation by obtaining some descriptive statistics. We started with the genotypes of the N = 697 unrelated individuals and quickly discovered that a large proportion (38.5%) of the 24,487 SNPs have a frequency of 1/(2N); that is, only one copy of the variant exists in the entire set of unrelated individuals, and fully 87.2% have a minor allele frequency less than 5.0%. Under usual circumstances, this collection of SNPs would have such a low polymorphic information content as to render ordinary linkage analysis powerless to identify genes involved in the quantitative phenotypes of interest. Fortunately for the family data set, the data simulators provided fully informative identity-by-descent (IBD) matrices for each pair of individuals at each gene. These IBD scores were given as 0, 0.5, and 1, denoting the sharing of 0, 1, or both genes identical by descent. Both methodological approaches reported here used these IBD matrices exclusively for the linkage analyses.
