Bayesian statistics for infection experiments

Abstract

To intervene cycles of food-borne pathogens in poultry new intervention methods need to be tested for their effectiveness. In this paper a statistical method is described that was applied to quantify the observed differences between test groups and control groups. Treated chickens and their controls were inoculated with several doses and were daily examined for the shedding of the tested pathogens. For these infection experiments with individually housed chickens and where binary data were available for each individual chicken a Bayesian analysis employing Markov Chain Monte Carlo (MCMC) was applied for the statistical analyses. The Cox’ proportional hazard reflected the typical features of the data, i. e. dependency, waitingtime structure and censoring. The outcomes of the analyses are two measures of difference in susceptibility between the feed groups. The first effect measure is a relative risk of being infected. The second is a difference in waiting time or a difference in inoculation dose to get a comparable proportion of infected animals