Hi Alfonso, thank you for the quick follow up! I should have been more clear in my description. Let's say my GLM for one session has three regressors (task_cond1, errors, rest) and another has four regressors (task_cond1, task_cond2, errors, rest). There are regressors that simply do not have events in session one, but I would like to include in the denoising for session 2. Is this possible? If you would like more information I can give more, it's a slightly more complicated design where we are trying to balance some regressors depending on the types of errors the subject makes, which differs by run. I can also upload an example mat file with the design if that helps. Thank you!

Originally posted by Alfonso Nieto-Castanon:

Hi,

I am not exactly sure what you mean. Different sessions always use different regressors, that is the normal scenario, they are just "named" in the same way. For example, in the Setup tab covariates (1st-level) you would enter as a "realignment" 1st-level covariate a separate file per subject and session containing the estimated position of that subject in the scanner during that session (e.g. a different rp_*.txt file for each run/session estimated during the realignment preprocessing step), and then during denoising you would indicate "realignment" as one of your potential confounding effects so that CONN will remove (again, separately for each subject and for each session) the effect of that covariate from the BOLD signal.

Hope this helps

Alfonso

Originally posted by hkeglovi:

Hello, Thank you for any assistance you can provide. We have a single-subject design where different sessions use different regressors which we would like to analyze. However, I am not sure how to go about setting up denoising, as it seems the code only ever looks at session 1 for what covariate names are possible (see screenshot of line in question from conn_process). Has anyone worked around this / have any suggestions for best denoising practices? Thank you in advance!