Hi François and everyone,
I am trying to run source statistics. (between subjects).
I have EEG data and I compute sources using a volumetric head model (openMEEG BEM).
For sources statistics, I tried t-tests with FDR correction and I performed the analysis just over a timelapse of 100 ms of interest. While there are quite high t-value when uncorrected, I found nothing with FDR correction (but wonder if it is not over corrected?).
I also tried ft_sources statistics but I get an error line 65:
"The input files do not contain full cortex maps, use function ft_freqstatistics instead". Is this error caused by the fact that my head model is volumetric?
What would you recommand for source statistics?
best,
Adeline
I found nothing with FDR correction (but wonder if it is not over corrected?).
If you are considering that repetitions across sources are multiple comparisons, then you have a very large number of tests considered in the FDR correction, and yes, it could lead to very strict corrections. In general, the more data you are testing at the same time, the stronger the effect of the correction for multiple comparisons.
Another approach is to work at the ROI level. The hypotheses of your experiment should include the ROIs in which you are expecting to observe an effect: use these ROIs as scouts and run your tests only on these scouts. Your problem of multiple comparisons will disappear.
I also tried ft_sources statistics but I get an error line 65:
"The input files do not contain full cortex maps, use function ft_freqstatistics instead". Is this error caused by the fact that my head model is volumetric?
It should be possible to use volume source files with ft_sourcestatistics, as long as you are not selecting the "cluster" correction.
I tried to fix the bug in this commit: Bugfix: FieldTrip source stat on files with no SurfaceFile · brainstorm-tools/brainstorm3@108f577 · GitHub
Please update Brainstorm and try again.
What would you recommand for source statistics?
See this tutorial:
https://neuroimage.usc.edu/brainstorm/Tutorials/Workflows
As usually, very helpful ! the bug is fixed 