Effects of marker information on sib-pair linkage analysis of a rare disease.

Model-free sib-pair linkage analysis was used to screen the GAW9-Problem 1 data set for evidence of linkage of a rare disease to any of 360 highly polymorphic marker loci. Negative regressions nominally significant at the alpha = 0.05 level were obtained for 44 markers; however all of these proved to be Type I errors. None of the four disease loci were detected by sib-pair linkage, which was not surprising, given the particular model and sampling scheme used to generate these data. Neither deleting parental marker genotypic information nor misspecifying marker allele frequency estimates substantially increased the Type I error rate. A two-stage testing procedure using a 10 or 20 cM map and a liberal first stage significance level gave the same overall results as a one-stage 2 cM map but required only about 42% or 22% as many markers, respectively.