Impact of Censored or Penalized Data in the Genetic Evaluation of Two Longevity Indicator Traits Using Random Regression Models in North American Angus Cattle

This study aimed to evaluate the impact of different proportions (i.e., 20%, 40%, 60% and 80%) of censored (CEN) or penalized (PEN) data in the prediction of breeding values (EBVs), genetic parameters, and computational efficiency for two longevity indicators (i.e., traditional and functional longevity; TL and FL, respectively). In addition, three different criteria were proposed for PEN: (1) assuming that all cows with censored records were culled one year after their last reported calving; (2) assuming that cows with censored records older than nine years were culled one year after their last reported calving, while censored (missing) records were kept for cows younger than nine years; and (3) assuming that cows with censored records older than nine years were culled one year after their last reported calving, while cows younger than nine years were culled two years after their last reported calving. All analyses were performed using random regression models based on fourth order Legendre orthogonal polynomials. The proportion of commonly selected animals and EBV correlations were calculated between the complete dataset (i.e., without censored or penalized data; COM) and all simulated proportions of CEN or PEN. The computational efficiency was evaluated based on the total computing time taken by each scenario to complete 150,000 Bayesian iterations. In summary, increasing the CEN proportion significantly (p-value 20%) of censored data has a negative impact in the genetic evaluation of longevity. The penalization criteria proposed in this study are useful for genetic evaluations of FL, but they are not recommended when analyzing TL.