Estimation of Interclass Correlation from Familial Data

In the analysis of familial data in order to quantify the degree of parent–child resemblance several different estimators of interclass correlation, the pairwise, sib‐mean and random‐sib methods, have been used. We compare these methods and propose a new estimator, the ensemble estimator. For the case where there are a fixed number of siblings per family, the pairwise estimator is shown to be equivalent to the maximum likelihood estimator. When there are a variable number of siblings per family, the pairwise and ensemble estimators are shown by Monte Carlo simulation to be preferable due to their smaller mean square errors. When the sib‐sib correlation is low, the pairwise estimator is more effective whereas at high values the ensemble estimator is more effective.