I have watched and read through numerous video tutorial concerning non-parametric thresholding and how to execute this in FSL. However, I am still unclear how this can actually be applied to the original data after the permutation tests have been run.
As I currently understand, non-parametric thresholding allows you to compute the distribution based on the data, given that parametric correction does not best control for false error rates. Taking this one step further, threshold free cluster enhancement is superior to cluster-based thresholding for permutations (assuming clusters are what you are after) because it accounts for some issues of defining the initial cluster threshold and smoothing. Between the Mumford Brain Stats tutorials and Kevin Choi’s tutorial, I understand how I can run the permutation tests with FSL randomise.
To apply this, how or where can one determine what the threshold might be in order to apply this to the original data? I assume it is no longer .05? Is there a way to export the data from the corrp_tstat file to create a histogram distribution?
Thanks in advance for helping clarify!