Skip to main content
. Author manuscript; available in PMC: 2012 Apr 19.
Published in final edited form as: Ann Appl Stat. 2011 Jun;5(2B):1159–1182. doi: 10.1214/11-AOAS476

Fig. 5.

Fig. 5

Comparison of voxel-wise predictive R2 (based on the validation data) of the log(1 + sqrt(X)) model (4.2) and the sqrt(X) model (4.1). The vertical axis shows the difference R2 of (4.2) −R2 of (4.1). The median improvement of model (4.2) is 5.5% for voxels where both models have a predictive R2 > 0.1.