Hi Lars,

Some thoughts on your questions below
Best
Alfonso
Originally posted by Lars Michels:
Yes, if you only define the "task" condition (just for consistency I am going to call your three types of blocks here "task", "control", and "baseline") then the PPI analyses (or gPPI when you only select the "task" condition) is going to give you the relative connectivity between the "task" blocks and the aggregated "control" and "baseline" blocks. For your design I would probably suggest to define at least the "task" and "control" conditions explicitly (in Setup.Conditions), and then select gPPI analyses and select both of these conditions. When you do this, if in the second-level results you only select: a) only the "task" condition, that will give you the relative connectivity between the "task" blocks and the "baseline" blocks; b) only the "control" condition, that will give you the relative connectivity between the "control" blocks and the "baseline" blocks; and c) both the "task" and "control" conditions (1 -1 contrast), that will you give you the relative connectivity between the "task" and "control" blocks. 

Also, if in addition you have pre- and post- scanning sessions, then you could simply create four conditions "task_pre", "control_pre", "task_post" and "control_post", and again select all four when performing gPPI analyses. 

Last, in case I have not mentioned this before, if you have relatively long blocks then both gPPI and weighted-GLM analyses (within-block connectivity estimates) are going to give you almost identical results, so if you find it simpler you could also simply define explicitly all of your conditions (e.g. "task_pre", "control_pre", "baseline_pre", "task_post", "control_post", and "baseline_post") and perform standard weighted-GLM analyses, which will allow you to look at the individual connectivity effects or at any arbitrary between-conditions contrast in your second-level results.


Yes, sorry about that, you cannot "move back" a first-level covariate into the conditions list (you need to re-enter the condition onset/duration vectors; sorry there is no simple fix, when moving a condition to the first-level covariate list only the associated regressor timeseries are kept so it is not possible to reconstruct the original onset/duration values from that timeseries)

And yes, just to be explicit, for your pre- and post- design, I would suggest to duplicate your conditions (i.e. "rest_pre", "task_pre", "control_pre", "rest_post", "task_post", and "control_post"). Then:

a) for psychophysiological interaction analyses you can run gPPI selecting the four "task_pre" "control_pre" "task_post" and "control_post" conditions (the not-explicitly-defined baseline blocks will act as baseline for the individual PPI effects in these analyses)

b) for Fair et al. style analyses you can run weighted-GLM selecting only the "rest_pre" and "rest_post" conditions.

c) for within-block connectivity estimates you can run weighted-GLM selecting the four "task_pre" "control_pre" "task_post" and "control_post" conditions

In all cases you would leave all of the four "effect of *" effects in the Denoising step 'confounding-effects' list.  

Hope this helps
Alfonso