The estimated variance–covariance matrices from (A) the best-fitting phenotypic random regression model of tolerance and (B) the full random regression animal model.

Note the best-fitting RRAM did not include the additive genetic by FEC interaction VA×FEC. The estimated variances (diagonal, marked with asterisks*), covariances (below diagonal, marked with a superscript a), and correlations (above diagonal, marked with a superscript b) are shown with standard errors in parentheses. We were primarily interested in whether the slope—and thus tolerance—variances were significant; these are highlighted in bold italics where significant (see text for details). The heterogeneous residuals allow the residual variance to change with increasing strongyle FEC. We allowed the residual variance in body weight (WT) to vary across four quartiles of FEC.