Comparison of ascertainment-bias correction schemes for pedigrees ascertained through multiple probands

Assessing the genetic causation of a disease, which is of prime importance in medical genetics, is usually done by analysing pedigree data. When gathering such data, it is often more practical to adopt a non-random sampling strategy. However, unless suitable corrections for non-random sampling are made at the time of data analysis, inferences may be grossly affected. For pedigree data ascertained through multiple probands, various correction schemes have been suggested, although the efficiencies of these schemes are unknown. This paper compares such schemes, using Monte Carlo simulation techniques, under a simple genetic model, for pedigrees of fixed sizes and structures and for probands of two types of relationship--parent-offspring, and a pair of siblings. It is found that gene frequencies are grossly overestimated and the penetrance value of heterozygotes slightly underestimated whether or not any correction for non-random sampling of pedigrees is made. Knowledge of the population value of the gene frequency improves the estimate of the penetrance parameter.