Analysis.correlation(evds, nIter=100)[source]

Computes pairwise correlations between the two data splits in half.

Reported as one minus a correlation over two to provide a dissimilarity measure between 0 and 1 as in svm().

Data is normalized by subtracting the mean across conditions (targets) per chunk per voxel.


evds (event-related mvpa dataset)

nIter (int, default: 100)

Number of random splits in half of the entire dataset.

A header and a results matrix with four columns:
  • iter: iteration number

  • stim1.cond: first condition

  • stim2.cond: second condition

  • subj_resp: one minus the correlation value over two

    (0: patterns identical, 1: patterns have nothing in common)