. Author manuscript; available in PMC: 2016 Jan 1.

Published in final edited form as: Neuroinformatics. 2015 Jan;13(1):7–18. doi: 10.1007/s12021-014-9222-9

Table 4.

Significant regions defined by bootstrapped analysis (Bilateral)

	Location^a			Size^a	Mean fractional anisotropy^a		Sensitivity/Specificity
Region	RL	AP	IS	Voxels (mm³)	PD	Control	Regional AUROC
Rectal Gyrus^Ŧ	1	−25	−21	1457	0.24+/−0.03	0.20+/−0.02	0.82
Middle Cingulate^Ŧ	1	18	38	1093	0.20+/−0.05	0.15+/−0.04	0.75
Left Putamen	18	−6	−3	1066	0.30+/−0.04	0.24+/−0.03	0.82
Right Putamen	−18	−10	−6	411	0.26+/−0.04	0.20+/−0.03	0.82
Left Thalamus	9	21	7	238	0.28+/−0.02	0.25+/−0.02	0.80
Mean of Regions^b					0.25+/−0.02	0.21+/−0.02	0.901

Raw results involved creating 42 separate renditions of the above tables. However, for the sake of clarity, we present location and size of regions derived from the full analysis (generated by taking the 90 % confidence interval from a map comprised of adding all 2,100 bootstrapped analyses). Similarly, mean and standard deviation of fractional anisotropy are shown based on the full bootstrapped analysis. Conversely, sensitivity and specificity (AUROC) is based on each of the 42 individual “left out” analysis

Mean value of each region, rather than mean of individual voxels (which would have weighted predictions to favor larger regions) was used to generate a “mean of regions” for the purposes of prediction