Comparison of models for the genetic evaluation of reproductive traits with censored data in Nellore cattle

Abstract

In typical genetic evaluation, often some females have missing records due to reproductive failure and due to voluntary and involuntary culling before the breeding season. These partially or unobserved phenotypes are known as censored records and their inclusion into genetic evaluations might lead to better inferences and breeding value predictions. Then, the objective was to compare prediction ability of models in which the phenotypic expression of age at the first calving (AFC) and days to calving (DC) were considered to be censored and uncensored in a Nellore cattle population. Age at first calving and days to calving were analyzed as following: uncensored animals (LM); penalization of 21 d (PLM); censored records simulated from truncated normal distributions (CLM); threshold-linear model in which censored records were handled as missing (TLM) or coded as the upper AFC/DC value within contemporary group (PTLM); and Weibull frailty hazard model (WM). Pearson correlations (PC), the percentage of the 10% best bulls in common (pTOP10%), accuracy of estimated breeding values (r), and a cross-validation scheme were performed. Heritability estimates for AFC were 0.18, 0.12, 0.12, 0.17, 0.14, and 0.07 for LM, PLM, CLM, TLM, PTLM, and WM, respectively. PC and pTOP10% were higher among linear models and smaller between these models and WM. The models provided similar r of sire breeding values. Heritability estimates for DC were 0.03, 0.08, 0.06, 0.02, 0.07, and 0.10 for LM, PLM, CLM, TLM, PTLM, and WM, respectively. Strongly associated predictions were observed in CLM, PLM, PTLM, and WM. The highest coincidence levels of sires in the TOP10% were between CLM, PLM, and PTLM. The r of sire breeding values obtained applying CLM, PLM, PTLM, and WM were similar and higher than those obtained with LM and TLM. In terms of prediction ability, WM, PLM, TLM, and PTLM showed similar prediction performance for AFC. On the other hand, CLM, PLM, PTLM, and WM showed the similar prediction ability for DC Therefore, these models would be recommended to perform genetic evaluation of age at first calving and days to calving in this Nellore population.