Calving ease is recorded as a binary trait, easy or difficult calving. Predicting breeding values as the probability of expressing easy calving requires using threshold models, a nonlinear statistical methodology with intensive computing requirements emphasized nowadays with big datasets. Linear models are computationally simpler and may be an alternative for predicting calving ease breeding values when threshold models are unfeasible; however, predictions from linear models are in the observed scale and not in the probability scale. Our objective was to propose transformations of estimated breeding values from the observed to the probability scale to obtain estimated breeding values from linear models analogous to those from threshold models. We tested our proposed transformations using American Angus beef cattle population data. Linear models were 5× faster to converge than threshold models. Spearman’s correlations between estimated breeding values in the probability scale from the threshold and linear models were 0.97 (maternal) and 0.99 (direct). Distributions of estimated breeding values overlapped, showing good agreement in shape and scale. Therefore, genetic trends from linear and threshold models were analogous. Our results suggest the genetic progress attained with linear and threshold models is similar.

Read the full article in the Journal of Animal Science.