I just wanted to check something about the output from your data2psiX function. Is it appropriate to compare the “raw” PSI output values from the function to 0 using a one-sample t-test (e.g., whether the mean PSI of a group sig. differs from 0)? By “raw” I mean not the z-transformed PSI values from the function’s permutation test. (Although, I guess a follow up question would be if the same test would be appropriate for the z-transformed values).
Thanks and all the best,
Hi James. For these kinds of spectral measures, the true underlying distributions are often unknown, plus the raw psi values can differ over subjects. So I would say it’s better to normalize to an empirical null hypothesis distribution first, which will transform the data into the same (standard deviation-based) scale. And then you could continue with group-level statistics from on those values.
Terrific, thanks for this! I’ll proceed with the t-test on the z-transformed values.
All the best,
Hi! Interesting discussion. Which course does this refer to? Thank you.
This is about a measure of synchronization that is sometimes used in neuroscience. The technical paper is here, FYI.
I don’t talk about this in any of my courses (yet…), but I wrote about it in a book on neural time series data analysis.
Hi Mike, Hi all,
I wonder if you could help me on something,
Z-scores should be computed for each subject, separately for given conditions of interest, say, for congruent and incongruent conditions, or we should just take the SD across all conditions for each subject?
Interesting question. My intuition is to use condition-specific standard deviations, just in case the variance is different across conditions.
It also has the advantage that the first and second statistical moments (mean and standard deviation) come from exactly the same sample data.