We next proceeded to examine variance component models that allowed for either genetic dominance or shared sibling environment (Tables 1 and 2, last two columns). Although the two components of variance have different interpretations, they lead to equivalent predictions of trait variances and covariances for most relative pairs, and thus are hard to distinguish statistically in most datasets [26], including ours. We interpreted the estimate of heritability from a model allowing genetic dominance as a liberal estimate for the impact of genes on each trait, and the estimate from a model allowing instead for a contribution of shared sibling environment to the variance as a more conservative estimate. In most cases, the true heritability will be intermediate between these liberal and conservative estimates.
