Logistic regression models for evaluation of physico-chemical composition of bovine milk “in natura”

Abstract

This study aimed to evaluation, through logistic regression model, the relationship between physical-chemistry composition of bovine milk in natura and probability of occurrence of mastitis in crossbred female cattle from Hostein, Gyr and Jersey breed. The result of the CMT (positive =1 and negative = 0) was used to study the probability of occurrence of mastitis that was modeled using logistic regression. The final model consists of fat, lactose and somatic cell count (SCC) was selected by the Stepwise procedure available in SAS^®, from the regressive variables farm, protein, degreased dry extract, fat, lactose, and SCC. It was observed that 53.86% of the animals had subclinical mastitis. The CCS, fat and lactose influenced the probability of incidence of mastitis and an increase in one unit of these variables is associated with an increase of 0.4% and 52.8% and reduction of 96.5%, respectively in the probability of incidence of mastitis. The CCS is the variable with the greatest impact on probability of mastitis, and the count of 600.000 cells/mL result in the 28% of probability of incidence of mastitis. The logistic regression model to quantify the impact from mastitis in the physical-chemistry composition of milk in natura in the dairy herd.