Table S1. Description of the full model and reduced models for testing different hypotheses on the relevance and interpretation of the genotype by induction treatment interaction (adapted from Yang, 2002; de la Mata and Zas, 2010). The reduced models constrain different elements of the family (G) and residual (R) variance-covariance structures by specifying in the randomstatement of proc mixed different types of covariance structures, and/or constraining different covariance parameters with the hold option of the parms statement. The number of parameters to be estimated is shown for each model (#par). All hypotheses are tested by comparing each reduced model with the full model, where the difference in two times the log-likelihood of these models is distributed as a one tailed χ2, with degrees of freedom given by the differences between the number of covariance parameter specifying the full model and the reduced model (Fry, 2004).