Because these are strong predictors of the semen quality indicators, their inclusion in the models results in gains in efficiency for the estimation of the association of interest by reducing the amount of unexplained random variability in the model.