The use of the log-linear model to evaluate the effects of three herd factors on Streptococcus agalactiae mastitis occurrence in California, 1977

Epidemiologic relationships among Streptococcus agalactiae infected dairy herds and herd size, location and California Dairy Herd Improvement Association (CDHIA) participation were examined using log-linear methodology. Dairies located in five out of the six California Bureau of Animal Health veterinary districts were studied. Data sources included the Dairy Cattle Data Base, which uniquely identifies each of the 2875 dairies in California, the 1977 statewide CDHIA year-end summary report and microbiological results from a 1977 statewide survey of bulk tank milk samples. These data were merged, fitted and the best model selected using likelihood-ratio statistics. The model included all four “main effects”, the six possible “1st-order effetss” and the “2nd-order effect” due to the interaction of herd size, location and CDHIA participation. Subsequently, a logit model was used to estimate the effect of the independent herd factor variables (size, location and CDHIA participation) on the log odds of the dependent S. agalactiae variable. This model required conditioning on the three-way relationship among the independent variables. medium-sized, non-CDHIA herds in district 2 (northcentral California) showed the highest expected odds (3.37) for S. agalactiae in bulk tank milk, while small, CDHIA herds in district 3 (northcentral coastal California) produced the lowest odds (0.39).