where in both models only one posture could be included at the time. The regression coefficient estimates bI and bII were of our main interest. Untransformed postures were included as independent variables in model II.
Using a stepwise selection procedure, EBVma (model I) and posture (model II) were included in the analyses by default. Other, independent variables could enter the model if they had an F statistic with P ≤ 0.05. After adding a new variable to the model, significance levels of other variables already included in the model had to remain ≤0.05, otherwise they were deleted from the model. No other variables (COV) were eventually included in model 1, whereas in model 2, mean piglet birth weight was always included and duration of parturition was only included in the analysis of duration of sitting.