Genetic evaluation for days to calving in Nellore heifers using Exponential and Gaussian Censored Bayesian models

•Days to calving play a key role in the overall profitability of Nellore cattle production.•Nellore females usually provide censored data for days to calving due to inconsistent data recording.•Censored Bayesian models were evaluated for genetic prediction of days to calving.•The Exponential model was not suitable for genetic evaluation of days to calving.•Censored Gaussian models are recommended for genetic evaluation of days to calving. Days to calving (DC) plays a key role in the overall profitability and long-term sustainability of tropically-adapted beef cattle breeds (e.g. Nellore), especially those raised under extensive production systems. However, some females usually do not have reliable phenotypic records because of late calving and even failures in its recording, which can be considered as censored data. Furthermore, DC tends to present an asymmetric distribution (such as Exponential) as it represents the time to an event occurrence, i.e., calving. Thus, we aimed to propose and evaluate the Exponential censored Bayesian models for genetic evaluation of DC in Nellore cattle as well as to perform comparison analysis (predictive ability and goodness-of-fit) with the conventional Gaussian models. Additionally, different scenarios were based on censoring definition according to the cut-off values: 321, 351 and 381 DC. The predictive ability was evaluated under a cross-validation approach, and the goodness-of-fit were performed using the Deviance Information Criterion (DIC). Gaussian models presented higher predictive ability and lower DIC (better goodness-of-fit) compared to Exponential models. Censored models provided higher predictive ability than uncensored models. The highest predictive performance was observed for Gaussian models, which also reported more realistic heritability estimates (ranging from 0.03 to 0.23). Additionally, Spearman correlations and coincidence rates were higher between different censored scenarios in Gaussian models. The inclusion of censored data into genetic evaluations could be recommended for Gaussian models, and the proposed Exponential model was not suitable for genetic evaluations of DC.