Add that minimum number of channels for defining cluster is 0 in documentation

[kerrencasper]

Proposed documentation enhancement

In mne.stats.permutation_cluster_test there is no way of defining the minimum number of neighbouring points that defines a cluster. This is a major error in the code, which makes the code invalid. This has been discussed by many of the developers for years, but nothing has been done (see issue #10604, #10604). Most people 1) don't know why this is a problem; 2) blindly trust the function. The least you can do if you are not ready to change it is to add to the documentation that this is a shortcoming of the function and that it should not be used until it has been solved. To exclude clusters with less than n number of channels (as has been discussed in the above-linked thread) is not a valid way forward as the t-values for these clusters have already been part of the t-distribution for defining the threshold for what can be considered significant.

[welcome]

Hello! 👋 Thanks for opening your first issue here! ❤️ We will try to get back to you soon. 🚴

[larsoner]

Hmm... I think our minimum number is effectively 1, which should make it so things are valid, right?