Hi tbn,

it may be helpful to consult with someone in your team with previosu experience in the method or with design matrix formulation. It is difficult to provide an extensive respose that cover all of your queries here. You may also want to consult to NBS user manual, which can be downlaoded from this site. It provides various example statistical designs that you can use as a template. 

Andrew

Originally posted by tbn:

Hi everyone,

I’m an intern and this is my first time working on functional connectivity analyses. I’m trying to prepare inputs for the NBS toolbox using gPPI matrices from the TMFC toolbox (https://github.com/Masharipov/TMFC_toolb...), and I’m a bit unsure if I’m doing it correctly. I’ve followed the NBS manual, but as a beginner, some parts aren’t very clear.

Here’s what I’ve done so far: I wrote a MATLAB script to process the subject-level ROI-to-ROI gPPI matrices and create NBS input files. I also wanted to include covariates like symptom scores (MADRS, STAI-S, STAI-T, SHAPS) and group (healthy controls, HC vs high symptom group, HS). Previously, I only used the TMFC toolbox, which outputs matrices for each subject for a specific contrast/condition (in my case it's just condition A), but with TMFC I can’t include covariates or group information.

The input files created:

1. 
   

   Subject matrices:

   
   
   
   
   • 
     

     Stored in a mat file (size: [nROIs x nROIs x nSubjects])

     
     
   
   
   • 
     

     Each slice along the 3rd dimension is a symmetric ROI-to-ROI connectivity matrix for one subject 

     
     
   
   
   
   
   
2. 
   

   Node labels:

   
   
   
   
   • 
     

     txt file which contains the list of ROIs corresponding to the rows/columns of the matrices.

     
     
   
   
   
   

    

   
   
   
3. 
   

   Design matrix:

   
   
   
   
   • HERE I AM REALLY UNSURE
   
   
   • 
     

     Size: [nSubjects x 7]

     
     
   
   
   • 
     

     Columns: intercept (a column of 1s) , HC dummy, HS dummy, MADRS z-score, STAI-S z-score, STAI-T z-score, SHAPS z-score.

     
     
   
   
   
   
   
4. 
   

   Contrasts:

   
   
   
   
   • 
     

     contrast_full_HS_HC.txt: HS vs HC comparison

     
     
   
   
   • 
     

     contrast_full_MADRS.txt, contrast_full_STAI_S.txt, contrast_full_STAI_T.txt, contrast_full_SHAPS.txt: each testing the association with a single covariate

     
     
   
   
   
   
   

My questions:

• 
  

  Does this workflow look correct for preparing NBS inputs from gPPI matrices?

  
  


• 
  

  Are all of the created files correct?

  
  


• 
  

  Is it okay to include covariates and group this way, or am I missing something important?

  
  


• What statistical test should I choose? I chose t-test so far ?


• I am using FDR, so how many permutations should I use? 

Any guidance or tips would be greatly appreciated, especially since I’m quite new to this. Thanks in advance!