. 2020 Jun 11;10(6):364. doi: 10.3390/brainsci10060364

Table 4.

Best-fitting models of age prediction and adjusted $R^{2}$ for each class and each harmonization technique for the three regression algorithms SVR, Random Forest and Lasso.

Harmonization Technique	Class	SVR	RF	Lasso
No harmonization	NC	$\hat{y} = - 0.02 y^{2} + 1.6 y - 3.25$	$\hat{y} = - 0.01 y^{2} + 1.16 y + 1.3$	$\hat{y} = - 0.02 y^{2} + 1.5 y - 2.6$
	NC	$R^{2} = 0.80$	$R^{2} = 0.82$	$R^{2} = 0.79$
	ASD	$\hat{y} = - 0.02 y^{2} + 1.54 y - 2.3$	$\hat{y} = - 0.006 y^{2} + 0.88 y + 3.8$	$\hat{y} = - 0.02 y^{2} + 1.7 y - 4.2$
	ASD	$R^{2} = 0.79$	$R^{2} = 0.81$	$R^{2} = 0.78$
Age covariate	NC	$\hat{y} = - 0.01 y^{2} + 1.4 y - 2.5$	$\hat{y} = 0.01 y^{2} + 0.65 y + 4.3$	$\hat{y} = - 0.01 y^{2} + 1.4 y - 1.8$
	NC	$R^{2} = 0.80$	$R^{2} = 0.83$	$R^{2} = 0.80$
	ASD	$\hat{y} = - 0.01 y^{2} + 1.4 y - 2.7$	$\hat{y} = 0.65 y + 5.7$	$\hat{y} = - 0.02 y^{2} + 1.6 y - 3.7$
	ASD	$R^{2} = 0.83$	$R^{2} = 0.80$	$R^{2} = 0.79$
No age covariate	NC	$\hat{y} = - 0.01 y^{2} + 0.93 y + 5.4$	$\hat{y} = - 0.01 y^{2} + 1.22 y + 2.44$	$\hat{y} = - 0.01 y^{2} + 0.82 y + 8.2$
	NC	$R^{2} = 0.48$	$R^{2} = 0.68$	$R^{2} = 0.40$
	ASD	$\hat{y} = 0.0005 y^{3} - 0.04 y^{2} + 1.3 y + 5.6$	$\hat{y} = - 0.0003 y^{3} - 0.03 y^{2} + 1.4 y + 2.4$	$\hat{y} = 0.0006 y^{3} - 0.04 y^{2} + 1.2 y + 6.9$
	ASD	$R^{2} = 0.33$	$R^{2} = 0.66$	$R^{2} = 0.30$