[FieldTrip] One-sided versus two-sided cluster statistics
Nina.Kahlbrock at uni-duesseldorf.de
Tue Feb 22 10:14:08 CET 2011
thank you very much for your reply!
I still have a question regarding your answer, though.
I propose that you calculate your p-values always one-sided. In fact, this
is what the FT permutation statistics functions also do. The difference
between a one- and a two-sided test is that you compare this one-sided
p-value either with your desired type-I error level (for a one-sided test)
or with half your desired type-I error level (for a two-sided test).
I think that the difference between 0.036 and 0.056 is due to the fact that
these are random quantities. If expect that, if you would increase
cfg.numrandomization to 100,000, you would find two p-values that are much
closer. In any case, the p-values are one-sided, and their calculation is
independent of the value that you choose for cfg.alpha.
I understand that cfg.alpha only gives the threshold for which of my
p-values is counted as being significant. However, I thought, changing
cfg.tail would change whether I am looking for positive and negative
clusters (cfg.tail = 0, two-sided test), or only positive or only negative
clusters (cfg.tail = 1 or -1, one-sided test).
The change in stat.prob from 0.036 to 0.056 described in my previous email
seems to depend on whether I set cfg.tail to 0 or 1. I've computed it
multiple times and witch cfg.tail = 0, stat.prob evolves around 0.03 and
with cfg.tail = 1 around 0.05.
Could you maybe explain this to me?
>From your explanation above I would assume that I can use cfg.tail = 0 and
cfg.alpha = 0.05 to end up with a mask including my significant clusters in
a one-sided test?
Thanks again for your help!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the fieldtrip