<div dir="ltr">More concretely, I would be very interested in a non-parameteric test: Permutation of train labels followed by cluster size correction, as for instance presented in the paper <a href="https://www.ncbi.nlm.nih.gov/pubmed/23041526">"Statistical inference and multiple testing correction in classification-based multi-voxel pattern analysis (MVPA): random permutations and cluster size control":</a><br><br>(A) Whole brain searchlight decoding using a permutation of the training label on a single-subject level. The whole procedure was repeated Mtimes (M ≥ 100) per subject for all N subjects, resulting in a pool of N × M single-subject chance accuracy maps. (B) One chance accuracy map was randomly selected per subject for 105 times. The selection of N maps (for N subjects) was averaged to one group accuracy map, resulting in a pool of 105 group chance accuracy maps. (C) Voxel-wise obtained empirical chance distributions, given the pool of 105 group chance accuracy maps. (D) The empirical chance distributions allowed the determination of a voxel-wise threshold which was used for a cluster search in the group chance accuracy maps.</div><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Oct 20, 2017 at 1:28 PM, Pegah Kassraian Fard <span dir="ltr"><<a href="mailto:pegahkf@gmail.com" target="_blank">pegahkf@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Hi,<div><br></div><div>I have been successfully running a searchlight analysis based on linear SVM for N different subjects - individually. Now I want to combine this information to plot searchlight results on the 2nd-level, i.e. group wise, correcting for the respective possible errors. How exactly can I do that? Is there an according pipeline available?</div><div><br></div><div>Best regards,</div><div>Pegah</div></div>
</blockquote></div><br></div>