Model comparison.

The table shows the log likelihoods (A, B) and the Akaike information criterion (D) comparing three models to experimental data. The growth curves of all nutrient conditions and the radial profiles of proliferation, death and ECM at a specific time were compared individually (A) as well as a whole (B, D). We also calculated the values according to the Bayesian information criterion (BIC), which led to the same ordering (not shown). Notice that the penalty terms in the AIC (2k) and BIC (k log(n), n: number of data points) play no role in the evaluation as the difference in the penalty terms are much (4–5 orders of magnitude) smaller than ln L. (Δ AIC = AIC(Model row-index)—AIC(Model column-index).)