Sample size calculations for Bayesian prediction of bovine viral-diarrhoea-virus infection in beef herds

Abstract

We used a Bayesian classification approach to predict the bovine viral-diarrhoea-virus infection status of a herd when the prevalence of persistently infected animals in such herds is very small (e.g. <1%). An example of the approach is presented using data on beef herds in Wyoming, USA. The approach uses past covariate information (serum-neutralization titres collected on animals in 16 herds) within a predictive model for classification of a future observable herd. Simulations to estimate misclassification probabilities for different misclassification costs and prevalences of infected herds can be used as a guide to the sample size needed for classification of a future herd.