Reproductive tract score, being the variable of interest, was initially used in
univariable models of days to calving and pregnancy outcome for the first to the fifth
breeding season, and of 24-day anoestrus in the first breeding season for each heifer
cohort as well as for the combined data. The individual effects of pre-breeding age, BW,
BCS and GL, and also the preceding season’s days to calving and CEI in the case of the
second to fifth calving seasons, were also estimated, whereafter the effect of RTS on the
outcome was adjusted for covariates that were significant (P < 0.05) predictors on their
own, using multivariable models. Year of birth was forced into all models of the
combined data, and AI bull was added as a random effect to the logistic regression
models of pregnancy failure during the first AI season. Confounding effects were deemed
to be present when inclusion of a potential confounding variable changed the regression
coefficients or odds ratios of other variables in the model by more than 30%, and such
confounding variables were kept in the models (Dohoo et al, 2003a). The fit of the logistic regression models of pregnancy failure during the first AI season was evaluated
using the Hosmer-Lemeshow goodness-of-fit test.