[Eeglablist] Questions about Permutation and Paired t test

[Zaeem Hadi]

Hello,
I have some conceptual queries. I have seen many recommendations to use the following method when trying to compare data statistically.
meanERSPcond1 = mean(erspdata{1,1},3);meanERSPcond2 = mean(erspdata{2,1},3);statcond({ meanERSPcond1 meanERSPcond2 }', 'method', 'permutation', 'paired','on','naccu',1000);
Where '3' is the dimension of averaging, which in my case is components in a single cluster.
First Q: Does the same process happens when we call std_erspplot??
Second Q: Why do we mean the data of each condition first and then make statistical comparisons? Shouldn't the subject/components be permuted across conditions?

Third Q: I have read the tutorial and article on cluster based permutation testing in Fieldtrip. Is EEGLAB's permutation testing similar to that?

I'd be grateful for the help or any reference material in this regard. 
Kind Regards,
Zaeem Hadi
Brain and Vestibular GroupImperial College London