All p values=1.0 in cluster based permutation tests

đŸ’¬ Support & Discussions
1

Hello,

I am running a number of cluster-based permutation tests, and each time, it results in p values all equal to 1.0. Would anyone be able to help pinpoint if I have set up my test wrong?

I am using repeated measures data from 50 participants, comparing between 2 conditions. Each participant’s data consists of 129 channels across 20 scales.

Below, I write the details for one of the tests.

Code for the statistical test:

w, clusters, clusters_pv, h0 = mne.stats.permutation_cluster_test(X, threshold=434, n_permutations=1000, stat_fun = wilcox_stat_fun, adjacency=adj_matrix)

Details on the parameters:

X is an array of the following dimensions: (2, 50, 129, 20)

Stat_fun: (wilcoxon signed rank test)

def wilcox_stat_fun(x, y, axis=0):
    return stats.wilcoxon(x,y,axis=axis)[0]

Adjacency: (accounting for 129 channels and 20 scales)

adj_matrix = mne.channels.find_ch_adjacency(raw.info, ch_type='eeg')[0]

adj_matrix = mne.stats.combine_adjacency(adj_matrix, 20)

MNE version: 1.5.0
Operating system: Windows 10

Thank you for your time!
-Diksha

2

Hi @diksha

Can you reproduce your issue using one of the MNE sample datasets? Something like

from mne.datasets.sample import data_path

fname = data_path / 'MEG' / 'sample' / 'sample_audvis_raw.fif' 
raw = mne.io.read_raw_fif(fname)
raw.pick("eeg")

# .... the code giving you trouble

If so, please share the code, which will make it easier for us to run it locally and help you figure out your issue