Dear experts,

I also have a question regarding repeated-participant design within the NBS toolbox. I followed the instructions as described in the FSL wiki and read some posts in this forum covering similar issues; however I am still not sure if I understood everything correctly. Could you possibly tell me if the following inputs are correct?

As a background: I want to determine differences in connectivity in two resting state conditions (RS1 and RS2). I have 57 participants, each participant completed both conditions.

• For the design matrix, I created a 114x58 matrix (see attachment). I modelled the conditions in the first columns, where RS1 is '-1' (first 57 rows in the first column) and RS2 is '+1' (rows 58-114 in the first column). The remaining 57 columns are modelling the mean effect of the subjects. Thus for the first participant, I included a '1' in the 1st and 58th row in the second column. For the second participant, I included a '1' in the 2nd and 59th row in the third column, and so on.
• Contrast vector: since I am interested in the effect of RS2, I made a 1 x 58 column vector containing a '1' only in the first column (see attachment). For the opposite effect, I created the same vector, just with a '-1' in the first column.
• Connectivity matrix: Since I use repeated design and use the AAL atlas (with 116 nodes), my connectivity matrix is a .mat file with the dimensions 116, 116, 114. In the third dimension of the matrix, 1-57 include connectivity matrices during the RS1 condition for each participant; 58-114 include connectivity matrices for RS2 condition for each participant.
•  Exchange blocks: for this, I created a 1x114 vector with repeating the sequence 1 to 57 twice (ie. 1,2,3,4,...,57,1, 2, 3,... 57)
• If I was to include a covariate of no interest (e.g. age), I would simply add the values as another column in the design matrix, correct? The contrast vector would remain the same though, right?

I also have some more general questions about the NBS toolbox:

• At the moment I use matrices which are not thresholded. Would it also be ok to use thresholded matrices (e.g. by using the threshold_proportional function from the brain connectivity toolbox?)
• Did I understand correctly that the treshold is is the t-statistic (if t-test is selected as statistical test) or F-statistic (if ANOVA is selected as statistical test)?
• In the NBSview window, sometimes several significant networks are shown. How does the toolbox determine that these networks are seperate/independent?

Thank you in advance for your answer,
Judi

Hi Judi,

Your design matrix looks correct. The contrast you have described is also correct. All the other comments you make are also correct.

Regarding a covariate: yes add another column to your design matrix for the covariate, but then remember to add another 0 to the end of your contrast vector to account for this extra column. If you don't add an extra 0, you will get an error.

1. You can threshold your matrices before NBS, although this needs to be done carefully since you are effectively forcing to zero connections in some subjects, which might be non-zero in other subjects.
2. Regarding the primary threshold in NBS, yes it is either the t-stat or F-stat.You can convert to a p-value if you prefer that.
3. Multiple significant networks are indeed possible. These represent distinct connected components. 

Andrew




Originally posted by juhu: