Backward stepwise elimination of nonsignificant interaction terms (P > 0.05 Wald’s test), starting with the 4-way interactions and ending with the 2-way interactions, was carried out. Thereafter, nonsignificant variables were eliminated to end with the best-fit final model consisting of significant variables and interaction terms except for the variables season and age at first delivery, where the whole effect is considered significant if an interaction term is significant and the single variable are not. For example, if 1 of the 2 season variables; that