Table 2.

Summary of Pearson r and R² results

	Task-state FC		Resting-state FC
	Pearson r	R2	Pearson r	R2
Overall	0.76	0.51	0.46	−0.29
Condition-wise	0.75	0.01✢	0.48	−1.71
Node-wise	0.77	0.48	0.46	−0.23
EMOTION:fear	0.76	0.52	0.41	−0.17
EMOTION:neut	0.76	0.51	0.38	−0.36
GAMBLING:win	0.87	0.73	0.52	−0.05
GAMBLING:loss	0.88	0.76	0.53	−0.05
LANGUAGE:story	0.67	0.05	0.31	−2.32
LANGUAGE:math	0.64	−0.16	0.27	−3.65
MOTOR:cue	0.68	0.43	0.53	0.21
MOTOR:lf	0.59	0.29	0.37	0.00
MOTOR:rf	0.56	0.25	0.36	−0.01
MOTOR:lh	0.57	0.26	0.37	−0.01
MOTOR:rh	0.55	0.23	0.34	−0.03
MOTOR:t	0.58	0.27	0.37	0.00
REASON.:rel	0.90	0.80	0.56	0.11
REASON.:match	0.90	0.81	0.57	0.13
SOCIAL:mental	0.90	0.80	0.51	−0.13
SOCIAL:rnd	0.89	0.78	0.53	−0.07
WM0bk:body	0.68	0.42	0.46	0.08
WM0bk:faces	0.72	0.48	0.40	0.02
WM0bk:places	0.75	0.53	0.51	0.16
WM0bk:tools	0.79	0.60	0.51	0.16
WM2bk:body	0.78	0.60	0.51	0.15
WM2bk:faces	0.78	0.59	0.42	0.04
WM2bk:places	0.72	0.47	0.47	0.11
WM2bk:tools	0.73	0.51	0.48	0.12

Pearson r and R² prediction accuracy assessment results are reported across task-state FC and resting-state FC. Results comparing all predicted to actual activation values at once are included (“overall”), as well as results comparing prediction accuracy across conditions for each node separately (“condition-wise”) and prediction accuracy across nodes for each condition separately (“node-wise”). See Table 1 for descriptions of Pearson r versus R² and condition-wise versus node-wise predictions. Node-wise comparisons for each task condition separately are included in the last 24 rows.

^✢Mean condition-wise task-state FC R² values were dragged down by regions with especially poor predictions because of mis-specified scales. Since mis-specified scales affect R² but not Pearson r values, poor scaling of predictions can be detected when R² predictions are much worse than Pearson r predictions. We found that 38 brain regions had R² values below −1.0, meaning the predictions were 100% worse than the average activation across all task conditions for those regions. Notably, these same regions all had Pearson r values >0 (mean r = 0.43), demonstrating this was primarily a scaling issue. In contrast, 186 (of 360) brain regions had R² >0.25, such that most brain regions exhibited overall accurate condition-wise predictions even when taking scale into account.