Analysis of Age at First Calving Dealing with Censored Data

Abstract Text: The objective was to compare prediction
ability of models in which the phenotypic expression of
age at the first calving (AFC) was considered to be cen-
sored and uncensored. AFC was analyzed as following:
Uncensored animals (LM); Penalization of 21 days (PLM);
Censored records simulated from truncated normal distri-
butions (CLM); Threshold-linear model in which censored
records were handled as missing (TLM) or coded as the up-
per AFC value within contemporary group (PTLM); and
Weibull frailty hazard model (WM). Pearson correlations
(r), the percentage of best bulls in common (TOP10%),
and a cross validation scheme were performed. Heri-
tability estimates were 0.18, 0.12, 0.12, 0.17, 0.14, and
0.07 for LM, PLM, CLM, TLM, PTLM, and WR, respec-
tively. TOP10% and r were higher among linear models
and smaller between these models and WM. WM produced
the best prediction ability, and could be then recommended
to perform the genetic evaluation of AFC.

Keywords: reproductive traits, Bayesian models, Nellore