The Effect of Nonrandom Sampling on Familial Correlations

Abstract

The effect on familial correlations of selecting families through a proband with an extreme value is considered. Nonrandom sampling of families causes substantial downward bias in correlations for the variable on which families were selected, even when probands are eliminated from the analysis. Familial correlations for variables not directly involved in the selection of families will usually not be affected by ascertainment bias.