In this work, we present the Bayesian approach as an alternative to frequentist analysis regarding correlated data of pH and N-NH3 in the Holstein cow rumen. It was observed that for pH and N-NH3 data, a posteriori estimates of coefficients of the regression models were significant, which was not observed for least-squares estimates. Thus, the Bayesian approach allowed inferences that were directly linked to the sampling of parameters of interest and statistical comparisons of non-linear functions of the estimated parameters.