Searchlight-based brain morphological classification analysis

[...]multivariate weight maps are harder to interpret than maps computed by independent analysis at each location. [...]classification for high dimensional data such as whole-brain data from a small sample size is prone to overfitting and thus requires dimensionality reduction (i.e. preselection of potentially informative voxel). The motivation of the searchlight approach is that it focuses the power of multivariate analysis on each region in turn, thus reducing dimensionality on the basis of the neuroscientific assumption that changes will cluster in contiguous brain regions.