Dear Alfonso,

We have a longitudinal study with a cross-over design where we are trying to demonstrate the effect of an intervention using resting state functional connectivity, in particular ROItoROI with the 32 network nodes included in the CONN toolbox. We have 3 timepoints (tp1, tp2, tp3) and 2 groups of patients (A, B).

The cross-over design means that the intervention and control periods were swapped between groups. That is, the intervention period for group A was between tp1 & tp2 (control period between tp3 & tp2) and for group B between tp3 & tp2 (control period between tp1 & tp2).

We have run the Group by Time interaction with some significant results. Now we would like to run a one sample paired T-test pooling together the data from both groups and comparing the intervention period vs the control period. For this, in the setup I would include the timeseries for tp1 to tp3 and the corresponding 3 conditions (tp1-tp3). For the second-level covariate, I´ll choose all subjects as created by default.

However, I´m not sure how this contrast could be specified when it comes to the conditions since this is a cross-over design and the timepoints included to calculate the difference in each period (intervention and control) will be the exact opposite between groups.
In sum, the one sample paired T-test we want to make would be:

intervention period [AB (tp2-tp1) & B (tp3-tp2)] vs control period [AB (tp3-tp2) & B (tp2-tp1)]

Would this comparison be possible to implement in CONN?

Please let me know if I didn´t make myself clear. Thank you in advance for your help.

Best wishes,
Noelia