help
help > QA variables in 2nd level MVPA analyis
Mar 13, 2018 02:03 PM | Clas Linnman
QA variables in 2nd level MVPA analyis
Hi,
I am analyzing a dataset of 292 subjects, scanned twice, under identical conditions.
I have followed the default preprocessing settings in Conn 17.f (no field maps)
I am interested in the influence of a subject variable (Xa during scan a, Xb during scan b), so I have set up the analysis as 292 subjects, 2 conditions per subject, and I entered Xa and Xb as 2nd level covariates.
Following Gabrielli 2015 (Brain connectomics predict response to treatment in social anxiety disorder) I did a MVPA (10 factors) and then perform a contrast on all 10 MVPA groups (eye 10), where I include Subjects, Xa, and Xb as subject effects, and the two scans as conditions. I then perform the between subject contrast 0 1 -1 (i.e. Xa – Xb), and the between conditions contrast 1 -1 (i.e. scan a – scan b). My understanding is that this will identify any brain regions where there is a correlation between the change in Xa vs Xb, to the change in rsFC metrics between condition a and condition b.
I will then use the resulting clusters as seeds in a post-hoc analysis.
The problem I have is that I find a cluster that looks like a CSF artifact (illustrated below at p=0.05) , and I was wondering what QA variables are suitable to include. Or if there is another way to get rid of this artifact.
Also, any comments on the approach, or suggestions on how many MVPA groups to use in the first level analysis, would be much appreciated!
Best
Clas
I am analyzing a dataset of 292 subjects, scanned twice, under identical conditions.
I have followed the default preprocessing settings in Conn 17.f (no field maps)
I am interested in the influence of a subject variable (Xa during scan a, Xb during scan b), so I have set up the analysis as 292 subjects, 2 conditions per subject, and I entered Xa and Xb as 2nd level covariates.
Following Gabrielli 2015 (Brain connectomics predict response to treatment in social anxiety disorder) I did a MVPA (10 factors) and then perform a contrast on all 10 MVPA groups (eye 10), where I include Subjects, Xa, and Xb as subject effects, and the two scans as conditions. I then perform the between subject contrast 0 1 -1 (i.e. Xa – Xb), and the between conditions contrast 1 -1 (i.e. scan a – scan b). My understanding is that this will identify any brain regions where there is a correlation between the change in Xa vs Xb, to the change in rsFC metrics between condition a and condition b.
I will then use the resulting clusters as seeds in a post-hoc analysis.
The problem I have is that I find a cluster that looks like a CSF artifact (illustrated below at p=0.05) , and I was wondering what QA variables are suitable to include. Or if there is another way to get rid of this artifact.
Also, any comments on the approach, or suggestions on how many MVPA groups to use in the first level analysis, would be much appreciated!
Best
Clas