Model selection with the data of Study 2.

MOI Models 2–9 were fitted to the pooled data of Study 2 [10]. We give estimates of model parameters with the 95% CI in parenthesis, and an asterisk indicates the lower and upper 95% CI limits coincide with the estimate parameter value. For each model we also provide the negative log likelihood (NLL), Akaike information criterion (AIC), the difference between a given model and the best-supported model in AIC (ΔAIC), and the Akaike Weight (AW). Overall, the best-supported model is Model 6, which combines the single mechanisms incorporated in Models 3 and 4. Of the models adding only one addition mechanism to the original Poisson model (Models 2–4), Model 3 leads to the greatest improvement in fit (i.e., it has the lowest NLL).