Dear all, my question regards statistics on time frequency decompositions. I have performed the morlet wavelets for my files and I want to threshold what is statiostcally significant. So I chose student t-test against baseline.It computes it however, he produced file does not have an option for viewing nor for setting a p value. Doing the same for the original signals give me the viewing and p value options.
Another question regarding the normalization by computing the z score. Is there a correlation of z score to p values???
Thanks for the feedback, Haris: this probably needs a fix in a future update of Brainstorm but I’ll let Francois confirm.
As for z-scores, there is the Chebyshev inequality that states that for any distribution of the data X as a random variable p( Z(X) > d ) < 1/d^2. Hence for d~4.5, you get p<0.05.
This is for one test: if you need to compensate for multiple hypotheses testing across the whole time-frequency plot, you can apply the Bonferonni correction p<1/(m x d^2), where m is the number of tests (e.g., the number of elements in a time-frequency plot). This is very conservative of course, but provides a little insight re: the value of Z with respect t statistical thresholding.
Overall, this all depends on the hypothesis you want to test, whether you are looking for an effect against baseline or against another condition.
Hi Haris,
Indeed, there is nothing yet in the interface to display the results of a t-test on time-frequency maps…
I will work on that at some point, but don’t expect those tools before a few months.
I’m sorry, the only option you have right now is to plot those values by yourself (right-click: File > Export to Matlab, the t a p values are in the pmap and tmap fields).
Francois