Table 4. Top-10 Layer-1 models from MEG ranked by variable importance.

ID	Family	Input	Feature	Variant	Importance	MAE
5	source activity	envelope	power	Ecat	0.97	7.65
4	source activity	signal	power	Pcat	0.96	7.62
7	source connectivity	envelope	covariance	$\alpha$	0.37	10.99
7	source connectivity	envelope	covariance	${\beta }_{low}$	0.36	11.37
4	source activity	signal	power	${\beta }_{low}$	0.29	8.79
5	source activity	envelope	power	${\beta }_{low}$	0.28	8.96
7	source connectivity	envelope	covariance	$\theta$	0.24	11.95
8	source connectivity	envelope	correlation	$\alpha$	0.21	10.99
8	source connectivity	envelope	correlation	${\beta }_{low}$	0.19	11.38
6	source connectivity	signal	covariance	${\beta }_{hi}$	0.19	12.13

Note. ID = mapping to rows from features. MAE = prediction performance of solo-models as in Figure 4.