Figure 7.

Average BIC of ${\mathcal{M}}_0$, ${\mathcal{M}}_1$, ${\mathcal{M}}_2$, and ${\mathcal{M}}_3$ when fitted on 100 independent data sets generated from ${\mathcal{M}}_3$ parameterized with N^* = 17, p^* = 0.27, q^* = 0.18 mV, σ^* = 0.06 mV, ${\tau }_D^{*}=202ms$, and ${\tau }_F^{*}=449ms$. ${\mathcal{M}}_3$ has the lowest average BIC compared to its submodels, showing that the parameters used to generate the data are indeed within the identifiability domain of ${\mathcal{M}}_3$. As a consequence, we can infer that ${\mathcal{M}}_3$ indeed provided the best fit to the data compared to its submodels, and that inferred parameters presented in Barri et al. (2016) are reliable. The facilitating nature of the synapse is illustrated by the fact that the BIC of ${\mathcal{M}}_2$ (the model with only STD and no STF) is substantially larger than the one of ${\mathcal{M}}_3$.