Hi Lars,

Some thoughts on your questions below
Best
Alfonso
Originally posted by Lars Michels:
Yes, you can simply define the control condition explicitly in Setup.Conditions with its associated onsets/durations (I am assuming here that the control condition is not implicitly define; i.e. if you have a ABABAB design with A your task condition and B your control condition then, for PPI analyses, it is not strictly necessary to explicitly define the B condition onsets/durations)

And regarding the PPI vs. gPPI, gPPI refers to Donald's generalization of PPI when there is more than one condition of interest, which simply models all condition effects and interactions simultaneously in a single model instead of using a separate model for each condition tested (in general the gPPI model is always going to be preferable over PPI when there is more than one condition of interest, and it reduces to exactly the same as PPI when there is only one condition).

The "effect of task*" effects are typically always entered in the denoising step, whether one wishes just look at the residual connectivity or one wishes to estimate task-related connectivity effects. This regression does not remove task-related connectivity differences, it only removes those connectivity differences that are directly due to a simple coactivation of two regions as a response to the task. See for example this post (http://www.nitrc.org/forum/message.php?m...) for a slightly more detailed explanation. 

Yes, for the same reasons as in the response above this regression only removes some portion of the connectivity differences (those that are simply due to coactivation as a response to the task) so the residual signal still contains task effects. In the framework of PPI, this regression removes the "main psycological" effect but it does not remove the "psychological by physiological interaction" terms (which are the ones we would typically be interested in when looking at task-related connectivity effects).

I would typically have both in the same conn project, and just run those as separate first-level analyses. For example, I would create three conditions "rest", "task", and "control". Then for analysis1 I would just use weighted-GLM (and look at the "rest" or "control" conditions), and for analysis2 I would just use gPPI (and select there the "task" and "control" conditions only; I am again assuming here that you do not have an ABABAB design -see this post http://www.nitrc.org/forum/message.php?m... for some discussion of why that makes a difference). 

Hope this helps
Alfonso