Non-parametric 2 sample t-test

💬 Support & Discussions
1

Hi experts,

My research involves the comparison between two conditions across two independent subject groups, which I believe to be a 2x2 research design.

Currently I have conducted 2 separate test, one for each subject group, to look at the difference between the conditions within each subject group.

This is done using mne.stats.permutation_t_test on the difference waves of the two conditions (mne.combine_evoked ([condition a, condition b], weights=[1,-1]), in each subject group.

From my understanding of the tutorials and manual, this seems to be a permutation paired t-test.

However I am also interested in finding the between subject group effect of condition a - b. I believe that I will need to carry out a permutation 2 sample t-test, which is not supported by mne.stats?

Another effect that I am interested is the interaction between subject group and condition, how can I test for that using non-parametric stats across the sensors using mne?

Lastly, I just want to verify my understanding that testing the difference between condition a-b and condition b-a is the same regardless of paired or two sample t-test?

Thank you sooo much!!!

Regards,
Vae

2

Hi @Heechberri - welcome!

I haven’t used the cluster permutation tests in quite some time (and am not an expert) but I don’t think that mne.stats.permutation_t_test can be assumed to be paired. For that I think there is mne.stats.permutation_cluster_1samp_test ?

3

Hihi:)

Thanks for the reply,

Just to confirm though the permutation tests in mne.stat are paired or 1 sample, is there a way to carry out a two sample test between 2 subject groups?

Thank you!!!

4

Hi @Heechberri

I totally mixed up in my head mne.stats.cluster_permutation_test with the mne.stats.permutation_t_test that you mentioned. You were right to consider the latter function paired, sorry for contradicting you!

per your 2nd question, maybe take a look at this mne tutorial that demonstrates a 2 samples permutation test and see if it helps you