The model selection procedure makes use of forward stepwise regression techniques to determine whether it is beneficial to include the sex-interaction terms, conditional on realizing a significant additive effect (p < 0.001). That is, the data are fitted to Equation 4 for a given marker, and if the add term is significant at the 0.001 significance level, then we attempt to add the sex*add term into the model. The sex*add term is retained in the model if the Bayesian information criterion (BIC) for this model is smaller than the BIC for Equation 4. If sex*add is included in the model as a result of this procedure, then we again use BIC to consider including the sex*dom term in the model.