Dear AFNI developers,
Is there currently an option for weighting group-level RFX analysis by some "quality" measure derived from the intra-subject level?
I'm referring to measures indexing intra-subject variability in a particular experimental condition (weighting a condition's Beta as function of subject-specific variance; FSL) or perhaps weighting Betas as function of tSNR in the functional run from which they are derived.
I repeatedly see particular brain areas associated with extremely high Betas that also have enormous variances associated with them on the intra-subject level. These betas are not trustworthy, but I currently have no systematic way of down-weighting these on the group level when using AFNI. In various AFNI docs I see that individual level variance is not propagated to the group analysis since “inter-subject variance is much larger than intra-subject variance” – couldn’t it be that in at least a few regions, inter-subject variance is large just because we propagate “nonsensical” betas to the group level?
Cheers,
Oori
