Table 2.

Performance of classification and regression models for each learning task (LT).

LT	Target	Target distribution	Method	Accuracy [%]	Sens. ♀[%]	Sens. ♂[%]
1	Gender		RIDGE	72.19 ± 2.94*	71.41 ± 5.46	73.02 ± 4.30
			SVM	71.82 ± 3.66*	70.93 ± 5.17	72.70 ± 6.00
			LASSO	71.33 ± 3.03*	69.49 ± 4.23	73.19 ± 4.88
			GBT	70.35 ± 3.36*	70.91 ± 5.05	69.72 ± 6.20
			RF	68.63 ± 5.24*	68.67 ± 5.95	68.58 ± 7.14
LT	Subset	Target	Target distribution	Method	RMSE	R2
2a	only ♀	TQ_distress (T0)		GBT	10.92 ± 0.68*	0.55 ± 0.04
				RF	11.38 ± 0.74*	0.51 ± 0.05
				LASSO	11.55 ± 0.70*	0.50 ± 0.04
				RIDGE	11.59 ± 0.63*	0.50 ± 0.04
				SVM	11.97 ± 0.51*	0.46 ± 0.03
2b	only ♂	TQ_distress (T0)		GBT	10.11 ± 1.12*	0.68 ± 0.06
				RF	10.58 ± 1.01*	0.65 ± 0.06
				LASSO	10.59 ± 0.98*	0.65 ± 0.05
				RIDGE	10.72 ± 1.05*	0.64 ± 0.06
				SVM	11.21 ± 1.02*	0.61 ± 0.06
3a	only ♀	ADSL_depression		LASSO	5.80 ± 0.73*	0.74 ± 0.06
		(T0)		GBT	5.88 ± 0.65*	0.74 ± 0.06
				RF	6.02 ± 0.62*	0.72 ± 0.05
				GBT	6.10 ± 0.65*	0.72 ± 0.06
				SVM	6.12 ± 0.72*	0.71 ± 0.06
3b	only ♂	ADSL_depression		LASSO	5.10 ± 0.38*	0.81 ± 0.03
		(T0)		RIDGE	5.14 ± 0.38*	0.80 ± 0.03
				GBT	5.16 ± 0.42*	0.80 ± 0.04
				SVM	5.29 ± 0.38*	0.79 ± 0.03
				RF	5.29 ± 0.46*	0.79 ± 0.04
4a	only ♀	Treatment effect		RF	9.55 ± 0.84	0.02 ± 0.09
		TQ_distress		RIDGE	9.58 ± 0.82	0.01 ± 0.09
		(T0 → T1)		GBT	9.59 ± 0.82	0.01 ± 0.08
				LASSO	9.61 ± 0.82	0.01 ± 0.08
				SVM	9.64 ± 0.83	0.00 ± 0.09
4b	only ♂	Treatment effect		RF	9.34 ± 1.31	0.04 ± 0.14
		TQ_distress		LASSO	9.41 ± 1.34	0.03 ± 0.14
		(T0 → T1)		RIDGE	9.43 ± 1.31	0.03 ± 0.14
				GBT	9.54 ± 1.32	0.00 ± 0.14
				SVM	9.55 ± 1.36	0.00 ± 0.14
5a	only ♀	Treatment effect		LASSO	8.21 ± 0.77	0.14 ± 0.09
		ADSL_depression		RIDGE	8.21 ± 0.80	0.14 ± 0.09
		(T0 → T1)		RF	8.23 ± 0.73	0.13 ± 0.08
				SVM	8.26 ± 0.84	0.13 ± 0.10
				GBT	8.26 ± 0.77	0.13 ± 0.09
5b	only ♂	Treatment effect		GBT	6.79 ± 0.73	0.17 ± 0.10
		ADSL_depression		RF	6.82 ± 0.68	0.16 ± 0.09
		(T0 → T1)		RIDGE	6.84 ± 0.73	0.16 ± 0.10
				LASSO	6.85 ± 0.71	0.15 ± 0.10
				SVM	7.15 ± 0.64	0.08 ± 0.09

Performance values are cross-validation mean ± SD. Algorithms are sorted by primary performance measure, i.e., accuracy for LT1 and RMSE for LT2-5. Asterisks signify models with significantly better cross-validation performance compared to the baseline w.r.t. the primary performance measure (α = 0.05). A bar chart (LT1) and histograms (LT2-5) illustrate the distribution of the target variables. For the numerical targets, median, median absolute deviation (MAD), and 95% bootstrap confidence interval (CI) are depicted.