Power comparison of generalizations of the mean test for affected sib pairs in case of incompletely informative markers

Previously applied test statistics for affected sib pairs (ASP) have been criticized for not fully exploiting the power of a given sample, because they do not account for incomplete informativity of an affected sib pair. To circumvent this problem, recently a new test had been proposed which weighs families proportional to their marker informativity. Here, the behavior of this new test is explored under scenarios considered to be relevant in the context of complex diseases. The results show that the new test does not represent an improvement. The asymptotic version of the new test tends to be markedly anticonservative even for larger sample sizes, whereas the permutation procedure proposed for the simulation of exact P values is at least questionable. Most importantly, the new test is less powerful than its competitors, especially in case of low marker informativity it was designed for.