Hello AFNI experts,
I’m attempting to ascertain the appropriate cluster threshold for FWE via 3dClustSim for a resting state analysis I’m performing. In the past, with task data, I have always used a residual error input data set for my 3dClustSim based on the rationale that I want to make my estimation based on an input that does not inherently model my design/anatomy(i.e. something which models noise). So I used the same approach with my resting state, acquiring a residual error output using the 3dDeconvolve -errts option with my preprocessed resting state image as my input.
The issue I’m having is in understanding exactly what this -errts output represents. The preprocessed image I use as my input is one which already has motion, csf and white matter regressed out. Thus, what I am left with is my model free baseline signal, which is the basis of my resting state signal I intend to measure. So I don’t know what noise signal the -errts image is giving me? Is it the case that my preprocessed image with physiological/noise already regressed out, and my -errts image are exactly the same? Since resting state is model free fluctuations in signal across the time series…which would be considered noise in the context of a task based GLM. And if this is the case, is it still suitable to use in 3dClustSim?
If not, what is the most appropriate input for 3dClustSim when ascertaining cluster thresholds for resting state data in the terms of connectivity analyses? Since ultimately I’m interested in the significance of clusters which demonstrate significant correlation to a seed, is it best to be putting a group level zmap through 3dDeconvolve and getting a residual for 3dClustSim? Or perhaps just run 3DClustSim on the group level zmap itself(this seems like cherry picking, to use my results to test the cluster threshold of results). I appreciate the help!
Sincerely,
Sam DeWitt