Spatiotemporal clustering permutation test with Representational Similarity Analysis (RSA)

💬 Support & Discussions
1

I would like to conduct a Representational Similarity Analysis (RSA) on MEG data. Essentially, it computes the Spearman correlation between a neural RDM and a model RDM for each subject, obtaining a rho value at each source and time point. I then want to perform spatiotemporal clustering to identify significant clusters across subjects where the correlation between the neural and model RDM is significant, followed by a permutation test to assess their statistical significance. I have been looking into MNE function mne.stats.spatio_temporal_cluster_test for spatiotemporal cluster analysis, but I’m not sure how to adapt it for RSA. Specifically, I’m encountering two problems:

  1. For the argument threshold: could I simply specify a rho value?
  2. stat_fun: how can I define the test statistic function to work with RSA and ensure that the permutation test properly shuffles the neural RDM values?

If spatio_temporal_cluster_test is not well-suited for this purpose as currently implemented, is there another function within MNE that provides a more flexible spatiotemporal clustering framework where I could specify RSA as the statistical test?

2

If you want to form clusters based on raw rho values instead of t-scores acress subjects, you can by using np.mean as stat_fun. However, I don’t quite see why you wouldn’t want to use the default ttest-acress-subjects in the case of RSA.

I’ve made an extensive tutorial on RSA which ends in doing cluster-based permutation tests for statistics, here: GitHub - wmvanvliet/neuroscience_tutorials: Some tutorials for students of neuroscience (look for “Representational Similarity Analysis using MNE-RSA” and “Statistics using a cluster-based permutation test”)

3 Likes
3

Hi Marijin, thank you so much for your response! I just read over these tutorials you sent over and found them to be extraordinarily helpful - this was exactly what I needed, thank you!

1 Like