Fig. 6: Statistical features underlying spatial predictions.
From: The spatial layout of antagonistic brain regions is explicable based on geometric principles
Upper panel: A the relationship between distance (great circle distance) and predictive accuracy for each of the tasks. Lower panel, spatial prediction using randomized task activation maps: B illustration of the first three Eigenmodes generated using Eigenstrapping20. These were used to generate 200 randomized versions of each task map while approximately preserving the spatial autocorrelation; these surrogate maps were then thresholded (<25% of vertices) and used to spatially predict out-of-mask vertices. C The similarity (Spearman’s ρ) between true and predicted out-of-mask vertices for the original task map (diamonds) and the randomized maps (dots).
