Description and model selection criteria.

aAll correlations were in infected subjects. bDIC (deviance information criteria) is a generalization of AIC in a Bayesian setting. DIC was not applicable (NA) in model 4, which assumed a correlation between more than two tests. cAIC (akaike information criteria) were used to evaluate goodness of model fit and to compare models. A difference in DIC or AIC of more than 10 indicated definite support to the model with the lower value, while a difference of between 5 and 10 was considered substantial, and less than 5 inconclusive.