help
help > RE: Longitudinal doubts
Sep 29, 2020 03:09 PM | Alfonso Nieto-Castanon - Boston University
RE: Longitudinal doubts
Hi Ines,
In a pre/post design (and assuming that in your second-level analysis you select both pre and post and specify a [-1 1] between-conditions contrast) the group-level analysis will include as dependent variable the differences in connectivity between the pre- and post- scans. In your results explorer view you may then select the 'plot effects' button to see, within each significant cluster, barplots displaying the average connectivity with the seed within each of the pre- and post- scans separately, which can be helpful when interpreting the group-level analysis results.
And regarding differences in MRI session parameters, yes, if there are any other potentially-confounding sources of differences in connectivity between pre- and post- scans (i.e. other than the "intervention" which you would like to attribute as the plausible cause of any significant effects), it is a good idea to explicitly encode those differences as a new second-level covariate and enter that explicitly as a control covariate in your second-level analyses in order to statistically control for those effects.
Hope this helps
Alfonso
Originally posted by Ines Del Cerro:
In a pre/post design (and assuming that in your second-level analysis you select both pre and post and specify a [-1 1] between-conditions contrast) the group-level analysis will include as dependent variable the differences in connectivity between the pre- and post- scans. In your results explorer view you may then select the 'plot effects' button to see, within each significant cluster, barplots displaying the average connectivity with the seed within each of the pre- and post- scans separately, which can be helpful when interpreting the group-level analysis results.
And regarding differences in MRI session parameters, yes, if there are any other potentially-confounding sources of differences in connectivity between pre- and post- scans (i.e. other than the "intervention" which you would like to attribute as the plausible cause of any significant effects), it is a good idea to explicitly encode those differences as a new second-level covariate and enter that explicitly as a control covariate in your second-level analyses in order to statistically control for those effects.
Hope this helps
Alfonso
Originally posted by Ines Del Cerro:
Hi,
I would like to add some comments to my previous post:
Dear CONN users,
I'm performing a longitudinal analysis with two temporal points (pre - post). I was wondering if CONN introduces any kind of covariate or regression variable at any point in order to correct possible differences caused by the MRI session (e.g. calibration of the scanner), or if it would be correct to use some covariate to take it into account.
We have also having some problems for the interpretation of our results. In a pre/post seed-to-voxel rsfMRI analysis, does the results correspond to the difference of the connectivity between both temporal points? Is that correct?
Thanks again,
Inés
I would like to add some comments to my previous post:
Dear CONN users,
I'm performing a longitudinal analysis with two temporal points (pre - post). I was wondering if CONN introduces any kind of covariate or regression variable at any point in order to correct possible differences caused by the MRI session (e.g. calibration of the scanner), or if it would be correct to use some covariate to take it into account.
We have also having some problems for the interpretation of our results. In a pre/post seed-to-voxel rsfMRI analysis, does the results correspond to the difference of the connectivity between both temporal points? Is that correct?
Thanks again,
Inés
Threaded View
| Title | Author | Date |
|---|---|---|
| Ines Del Cerro | Sep 16, 2020 | |
| Alfonso Nieto-Castanon | Sep 29, 2020 | |