We compared the robustness of affected-sib-pair (ASP) tests for multip
le-affected sibships. Forming all possible pairs increases the type I
errors only slightly whereas the most used weighting procedures decrea
se the efficiency of the tests. Another weighting procedure accounting
for the reduction of variance of the weighted identical by descent (I
BD) information appears robust. Missing parental marker data leads to
a decrease of type I errors in all cases. (C) 1997 Wiley-Liss, Inc.