Assessment of the relationship between the postpartum diseases susceptibility and the bovine monocyte subsets via Bayesian logistic regression, under various prior distributions

Abstract

Postpartum diseases (PD) in dairy cows cause serious concerns about economic losses worldwide. This study intended to investigate the relationship between PD susceptibility and counts of monocyte subgroup cells (MCC), in the blood samples taken from 27 German Holstein cows 42 and 14 days before the expected calving by adopting the Bayesian approach. The paper also aimed to discuss the prior selection problem in the Bayesian approach and to reveal the parameter estimation difference based on the available data. The parameters were estimated according to the models established at two different time points with eight different prior distributions. As a result of the study, all the models revealed strong evidence that cows with PD, compared to healthy cows, had a higher increase in MCC counts on Day 14. There was no difference between the models according to their WAIC and LOO values. In terms of the parameter estimates, the models produced identical results; however, the models with noninformative priors presented strong evidence for the absence of effects by Bayes factor but, provided evidence for the existence of the effect according to the credible interval. The models with weakly informative and shrinkage priors provided strong evidence for the presence of the effect. The findings suggest that MCC can be considered to serve as a prospective indicator for early detection of PD.