Fig. 4

Correlations between the population parameters in Model (2) allow for a reduction of its fixed effects. A Correlation heatmap (Spearman’s \(\rho ^2\)) of the 8 unidentifiable population parameters in the 50 runs of SAEM for the initial model. For each pair of unidentifiable population parameters, the heatmap displays the color-coded value of \(\rho ^2\). B: Nonlinear correlation between \(\delta _{SC}^{pop}\) and \(\rho _C^{pop}\). C: Linear correlation between \(\delta _{CB}^{pop}\) and \(\rho _C^{pop}\). (B, C Displayed are the estimated population parameter values over the 50 runs of SAEM, color-coded by likelihood.) D Estimated parameter values in the 36 convergent runs of SAEM for Model (7), with reduced \(\delta _{SC}^{pop}\) and \(\delta _{CB}^{pop}\). Displayed are the distributions of estimated parameter values, normalized by their median