Correcting for ascertainment bias in the COGA data set.

The effects of several corrections for ascertainment bias on a linkage result in the COGA data are examined. A correction that models the complete ascertainment protocol markedly reduces the evidence for linkage. Different partial corrections increase the linkage signal and recover good estimates of the population trait prevalence, but do not represent adequately the complexity of the COGA ascertainment scheme. Evidently the present size of the COGA data set cannot support a complete correction for ascertainment bias, but an effective partial correction can and should be implemented.