Dear CONN community,

I am working with a within-patient cross-over design and would appreciate your thoughts on my strategy to analyse functional connectivity in CONN.

Design
Each participant underwent 4 resting-state fMRI sessions:

1. 
   

   Baseline

   
   
   
2. 
   

   Treatment or placebo phase

   
   
   
3. 
   

   Baseline (after wash-out)

   
   
   
4. 
   

   The alternate phase (placebo or treatment)

   
   
   

The order of treatment/placebo was counterbalanced.

Goal
Previous work shows a robust treatment effect on functional connectivity (FC) in two predefined networks (seed-based).
I would like to test:

• 
  

  within-subject effects of treatment vs. placebo

  
  


• 
  

  patient-level effects (between-subject variability)

  
  

Because CONN currently does not implement mixed-effects linear models across multiple sessions, my idea was to:

• 
  

  Compute FC in the predefined networks and set up a contrast that averages the effect across all four sessions (e.g. baseline, treatment/placebo, baseline, treatment/placebo weighted 1/4 each).

  
  


• 
  

  Extract the resulting subject-level FC values.

  
  


• 
  

  Run a mixed-effects model in R, using participant ID and timepoint as random factors and treatment as the fixed effect.

  
  

Would you consider this a valid workflow, or would you recommend a different modelling strategy within CONN (e.g. a different second-level design or custom contrasts) before exporting the data?

Many thanks in advance.

Best regards,
Vanessa