Hi all!
I need to run a statistical analysis across four independent groups for some network measure I obtained with the brain connectivity toolbox (degree, clustering coefficient, etc.) from EEG data. I am dealing with the multiple comparison issue. Since I am using a high density cap, nothing survives the Bonferroni correction (and variants), nor the FDR.
I am trying to perform a cluster-based (montecarlo) analysis as implemented in Fieldtrip, but the toolbox needs the input to be organized as a time-frequency structure.
I tried to rebuild a similar structure (degree values instead of power spectrum values), but it seems that I am being unsuccessful.
Indeed: if I run a "cfg.statistic = indepsamplesF" across the groups I get NaNs in the probability output matrix. If I just perform a "cfg.statistic = indepsamplesT" between two random groups, I get a couple of channels or so with p
Hi Ramtin! It's been three years since you've written your question and I am just now struggling with the same problem you've described. When I run an "indepsamplesF" Test I get NaNs for my stats and probabilities, running a t-Test gave me an interpretable output. I was wondering if you were able to find the error? Thank you so much in advance for your help!
Hi Enya Weidner !
I did not explore that approach any further, hence I cannot give you any hint on that unfortunately.
However, a possible workaround to this could be to first use the Network Based Statistics (NBS) toolbox to assess any differential subnetworks between groups, and then obtain the nodal measures within the resulting subnetworks. This would be a safe way to reduce the number of tests!
Thank you for replying! I solved the issue, I had NaNs in my data which caused the F-test function to generate only NaNs.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix