Table 7.

Variable selection accuracy measured by false positive rate (FPR ± SE) across dimensional regimes under a nonlinear simulation framework. Lower values indicate improved ability to exclude irrelevant variables.

Models	Low ($p=6$)	Medium ($p=170$)	High ($p=2000$)
GBM	0.00 (0.00)	0.62 (0.169)	0.04 (0.031)
MEGB	0.00 (0.00)	0.66 (0.190)	0.06 (0.032)
GPBoost	0.00 (0.00)	1.14 (0.182)	0.07 (0.023)
MERF	0.00 (0.00)	0.57 (0.341)	0.06 (0.028)
REEMForest	0.00 (0.00)	0.44 (0.290)	0.07 (0.024)
RF	0.00 (0.00)	0.44 (0.290)	0.05 (0.000)
LMER	0.00 (0.00)	1.19 (0.00)	*
glmmlasso	0.00 (0.00)	0.85 (0.298)	*