Dear Lukas

Yes, all those options (including filtering, RegBP, linear detrending, etc.) look perfectly fine (just remember to also run some of the "denoising" quality control plots in order to better evaluate whether the level of denoising achieve by those default/recommended settings seems appropriate to your data). 

And regarding removing the conditions of no interest and keeping all those as a single implicit baseline, there is nothing wrong in principle (from a theoretical/mathematical perspective) with that approach, the only things to keep in mind is that by doing so you are assuming that difference in BOLD responses between those conditions should not to be excluded from connectivity estimates (meaning that those differences will be considered "baseline" fluctuations possibly part of the estimated baseline connectivity between regions). 

Best

Alfonso

Originally posted by Lukas Van Oudenhove:

Dear Alfonso,

I am running a gPPI analysis on a task-based dataset with four conditions of interest (tasting sips of 3 sweet substances and water), with swallow_rinse and ratings conditions of no interest.

I imported my data in the Setup tab, and got rid of my conditions of no interest (which may not have been the best choice, as I may need to remove their effect in the denoising step?). This means that I treat everything outside my conditions of interest as implicit baseline, do you think this is ok?

I am now at the denoising step, and based on previous forum posts of this topic, I understand that 1) adding the effects of the conditions of interest to the confounds should not matter much (although it feels counterintuitive to me, so would remove them), and 2) a high pass filter [0.008 Inf] may be better than the default band pass filter. Could you please confirm these are the right choices, and advise me on other choises (filter regressors or not, use RegBP, linear detrending, despiking)?

Screenshot of my current denoising tab attached.

Thanks very much in advance.

Best wishes,

Lukas