Modelos hierárquicos aplicados a dados biológicos de contagem

Abstract

A database from an experiment with two groups of 15 Nelore cows (Bos Tauros indicus) was analyzed. One group received a control supplementation (treatment 1) and the other received a supplementation with inclusion of long chain fatty acids salts (treatment 2). The data has the amount of aspirated and viable oocytes in a total of collects ranging from 2 to 8 aspirations per cow in a 120 days experiment, approximately. The objective of this study was to evaluate possible differences between treatments in relation to oocytes quantity and viability produced, as well as to identify best fit models through the Frequent, Bayesian and Bayesiana Longitudinal approaches. We used some hierarchical models and probability distributions. It was possible to conclude that there was no significant difference between treatments by the three approaches. In addition, by the Frequentist analysis, the best adjustments were made, respectively, by Binomial-Negative, Beta-Binomial and Poisson-Generalized models. In Bayesian analysis, the best adjustments were the Binomial, Beta-Binomial, Binomial-Negative, Poisson-Generalized and Binomial-Poisson-Generalized models. On the other hand, by the Bayesian Longitudinal approach, all models were adequate.