Figure_S1.tif (303.47 kB)
Figure S1 - A Two-Stage Cascade Model of BOLD Responses in Human Visual Cortex
figure
posted on 2013-05-30, 02:30 authored by Kendrick N. Kay, Jonathan Winawer, Ariel Rokem, Aviv Mezer, Brian A. WandellModel selection using alternative metrics and different threshold levels. As an alternative to cross-validation, we evaluated the accuracy of the CC, DN, CSS, and SOC models using Akaike's information criterion (AIC) and Bayesian information criterion (BIC). Here we plot model accuracy as a function of the number of voxels considered (voxels are selected based on GLM cross-validation accuracy; see Methods). Lines indicate the median accuracy across voxels in a given visual field map, and shaded regions indicate standard error (68% confidence intervals). Trends in model performance are consistent across metrics and are robust with respect to the number of voxels used in the model comparison.
(TIF)