The aim of this study was to evaluate the feed conversion (CA) by Bayesian inference in bivariate considering analyzes in real and simulated data. Different animal species experiments conducted at the Universidade Federal de Viçosa, state of Minas Gerais, Brazil are used. The proposed model proved to be appropriate once it enabled the detection of significant differences between levels of factors not detected by frequentist procedures with traditional ANOVA, especially in small samples. In the experiment with quails, it became clear that the birds' brute protein levels were 23% and 29%, respectively, for males and females, which presented better CA, 2.83±0.03 and 2.66±0.03, respectively. In the experiment with chickens, the group without additive antibiotic, including 0.02% extract natural esters promoted the best CA (1.72±0.01) and in general antibiotic absence of esters natural promoted 1.63±0.02 of the CA. In goats, it has been found that feeding milk from cows or goats also promotes better CA, respectively, in groups milked up to 60 and 90 days, being 1.29±0.14 and 1.79±0.11, suggesting that suckling done until 60 days. Pigs fed the highest level of metabolizable energy and aminoacids promoted the best CA (2.86±0.07) compared to a diet with lower nutritional level. But the use of enzymes in the diet with lower energy level and amino acid provided intermediate result (2.90±007). In cattle, it was observed that the use of 1% concentrate diet, CA, promotes a better estimate of 7.33±0.35 between Nellore and this promotion would be 7.40±0.58 between the cross breeds using 2% concentrate diet.
animal production; inference; nutritional performance; MCMC; multivariate