Hi gimme community,

I am running a resting-state analysis on timeseries extracted from 7 ROIS from the Hammers atlas (IC_ant, IC_post, AC, PC, IFG, Thalamus, Accumbens) for 55 subjects.
Data Pipeline:
1) minimal preprocessing with fMRIPrep
2) smoothing (FWHM=4mm) and denoising using CONN (some confounds removal)
3) extract timeseries from ROIs with nilearn
4) run gimme (method='Fast_greedy') on denoised timeseries

(Step2 could be done in python too, but this is the state of the matter).

After running gimme, looking at the IndivPathEstimates.csv file,
I get some high beta values (absolute value larger than 1.2): examples
# A tibble: 11 x 1
beta
1 -2.00
2 6.90
3 -5.17
4 -1.31
5 1.31
6 -1.23
7 1.30
8 -3.55
9 17.4
10 5.52
11 -60.4

from as low as -60.4 up to 17.4!

Does anybody know where could these high values come from / what could be the reason they are so high?

Thank you in advance for your comments / help

gio