Response bias in epidemiologic studies can occur if affected individua
ls are more (or less) likely to participate in a survey than their una
ffected counterparts. To examine the effect of response bias in the co
ntext of a family study, we conducted segregation and linkage analysis
in all 1,000 individuals in the Problem 2 data set, and in two differ
ent 65% samples: one sample consisting of 648 randomly selected indivi
duals, and the other sample nonrandomly constructed so that individual
s with high levels of Q1 were oversampled. In this simulation the abil
ity to detect major genes for Q1-Q4 in segregation analysis and to lin
k these putative major genes to genetic markers in linkage analysis wa
s not markedly different between the 65% random and the 65% enriched s
amples. (C) 1995 Wiley-Liss, Inc.