Hello Andrew,

thank you so much for your swift reply! 

Originally, we wanted to test first whether the change in connectome is even significant for pre and post, that's why we were using both data points. Do you know if it is even possible to test the second hypothesis (time*neurotiscism) without using the difference matrices? If not, we'll probably switch methods for all hypotheses.

Thanks again for your answer, it helped a lot already!

Best, 

Eva

Originally posted by Andrew Zalesky:

Hi Eva,

For the Response design matrix, the column for group needs to be removed when modelling the interaction between group and time. This is because group (responder vs non-responder) does not change between the two time points. If you don't remove the group column, a rank deficient warning will be given. 

Alternative modelling strategy: the difference in connectomes can be computed between pre and post and then the difference connectome can be inputted to NBS for each individual. In other works, each individual is associated with only one connectivity matrix (pre minus post connectivity). A t-test can then be used to determine whether the pre minus post difference in connectivity associates with response status. 

For the neuroticism hypothesis, I recommend using the difference connectome (as I describe above). The design matrix in this case will have a column of 1's and a column for the neuroticism variable. This would test whether the change in connecitvity links to baseline neuroticism. 

Best, 

Andrew

 

Originally posted by Eva Herter:

Hello everyone,

I have a question regarding the NBS-Toolbox. We collected MRI data from patients at two time points and now want to run two 2x2 repeated measure ANOVAs, with the two time points (pre and post; binary), therapy response (responder, non-responder, binary) and neuroticism (only pre; metric) as factors. For the first hypothesis we want to investigate whether changes in the connectome are associated with therapy response.

I have attached the DesignMatrix, contrasts and exchange blocks below, just to make sure we coded everything correctly.

For the second hypothesis we wanted to see whether changes in the connectome are associated with neuroticism, which we only measured at the first time point (since it’s a stable variable). We were wondering how you’re supposed to code the Design-Matrix for that. Should we put 0s in for the second timepoint and calculate the interaction that way? Or should we put the same values for Neuroticism in both time points? Is there a difference in desgin between binary and metric variables? The second sheet includes an option for the second hypothesis, maybe someone knows the correct coding.

I would really appreciate the help!

Eva