Smith, Rebecca2012-12-172016-12-302011-08-31bibid: 7955532https://hdl.handle.net/1813/30713Mycobacterium avium subsp. paratuberculosis (MAP) causes infections in ruminants characterized by long latent periods, imperfect diagnostic tests, and production effects sufficient to result in early culling. In order to optimize control of this pathogen, the exact effects of both the infection and the control strategies must be understood. The goal of this dissertation is to statistically analyze various aspects of MAP in dairy herds, including the production effects of infection, its distribution in the dairy herd environment, and the transmission rates of animals shedding MAP. For this purpose, longitudinal data sets from commercial dairy herds have been analyzed, allowing for a more thorough understanding of MAP in typical farms. Cows in the high-shedding category (>30 cfu/g of MAP in feces) were found to produce approximately 4 kg less milk per day, on average, and to have higher culling rates and lower calving rates than non-shedding cows. In addition, the number of high-shedding animals in a pen was positively correlated with the amount of MAP cultured from the environment in that pen. In contrast, low-shedding cows were found to have higher culling rates than nonshedding cows, but no significant difference in calving rates or milk production. The average amount of fecal shedding in the herd was found to be predictive of both the odds of MAP being found in environmental samples and the amount of MAP in those samples, but environmental sampling was not found to be a sensitive herd-level diagnostic test. These results will enable optimization of economic models for MAP control by providing quantitative estimates of the effect of MAP on commercial dairy farms. In addition, it was found that reversible-jump Markov Chain Monte Carlo models are unable to estimate transmission rates for MAP using current longitudinal data sets, due to the large amount of missing data.en-USstatistical modelingMycobacterium avium paratuberculosislongitudinal datadissertation or thesis