Hierarchical Bayesian models for robust estimation and censored data analysis in animal breeding.

摘要: Data strongly influenced by factors not accounted for the statistical model can bias estimates of genetic parameters and values. Moreover, several traits economic importance do follow a normal distribution or have censored data. The objective this study is to describe illustrate application hierarchical Bayesian models detection muting outliers analysis First, traditional specification animal in stages presented under approach normally distributed uncensored Then, extended introducing an independent weighting variable, which allows thick tail residual densities from Normal/independent family. Finally, cover data analysis, basic inclusion variable with truncated based on lower limit observed value trait at evaluation time, those animals that yet completed their reproductive life time.