Dear List,
My multivariate analysis (using support vector machine, svm) showed a neural network that classifies one condition more often than the other. Currently, I'd like to do a PPI analysis on this network. For the multivariate analysis, native-space unsmoothed data were used for 1st-level and then normalised and smoothed at 2nd-level.
Would it be best if do a GLM of the unsmoothed native-space data at the 1st level to extract BOLD signal and then generate the PPI interaction term for each subject?
Thanks very much!