Dear PRoNTo-Experts,
I'm dealing with uncertainty, whether my approach can be done by means
of PRoNTo.
My general goal is to predict subject's behavioral score (one value per
subject) from subjects brain activation.
I already did the covariate analysis, which revealed several clusters
correlated with the behavioral data. But since the correlation was not
that strong I wanted to run the multivariate analysis in the hope to
improve the predictive value of the brain data (I hope that multivariate
approach will identify patterns that are better correlated with the
behavioral score than just single voxels). For this I tryed to design
the regression analysis in PRoNTo, but I'm unsure, whether I'm doing right.
I've just started, but this is what I already did:
1) created the Group 'Group1' (is it correct to put all the runs into
one group or should I better create one group per run?)
2) clicked the tick box 'Scans' under the 'Subjects/Scans' panel (this
should allow regression in further steps)
3) created a new modality (fMRI)
4) now I have to select files, but I'm unsure which ones (there are many
possibilities: raw images after preprocessing or betta/con images after
1st level analysis; scans just from 1 run or from all runs; individual
scans or scans from all subjects). In the regression targets one value
per scan is expected. So it might be, that I have to select just one
scan for each subject.
I would appreciate any help regarding the pipeline and especially
regarding the step (4).
|