Practical identifiability in the frame of nonlinear mixed effects models: the example of the in vitro erythropoiesis

Table 1 A comparison of our models along reduction

Model	\(-2\,log( \widehat{{\mathcal {L}}} )\)	Convergent runs	Fixed effects	Random effects	\(\eta\)-shrinkage
Model	\(-2\,log( \widehat{{\mathcal {L}}} )\)	Convergent runs	Fixed effects	Random effects	\(s_{\rho _S}\)	\(s_{\delta _{SC}}\)	\(s_{\rho _C}\)	\(s_{\delta _{CB}}\)	\(s_{\rho _B}\)
(2)	1761	45	5	5	3	31	77	63	39
(7)	1761	36	3	5	3	28	66	40	62
(S1)	1761	36	3	4	3	28	–	49	44
(S2)	1761	22	3	4	3	29	63	32	–
(8)	1761	25	3	3	3	28	–	21	–

The table displays the optimal likelihood, the number of convergent runs (as selected by Akaike’s weights over 50 SAEM runs), the number of fixed and random effects, as well as the average \(\eta\)-shrinkage for each parameter (expressed as a percentage of the population variance) over the convergent runs for our models. For each model, we give the reference of the equation where it is defined.

ISSN: 1471-2105