Measles SEIR model parameter estimates and model selection using a Poisson sampling model.

Maximum likelihood (ML) parameter estimates for both models and two time series of the number of reported measles cases in two different cities: London and Birmingham. The sampling model for the observation error of the counts is the Poisson distribution. The letter denotes the number of model parameters in each case. denotes the value negative log-likelihood function evaluated at the ML estimates. The AIC and BIC scores for each model vs. data set combination are also reported. The model selection decision rule is to pick the model with lowest information criterion value. Accordingly, the LHD model seems to be the best choice in both data sets. Confidence intervals for and are shown in parentheses for the best model for each locality.