Hi,

I am evaluating a large dataset of rsFC (n=293), where subjects were scanned using identical parameters on 2 occasion, A and B, between 20 and 500 days apart.
I am interested in the change in rsFC between the two scans as it relates to change in anxiety. There is a nice spread in the anxiety scores, ranging from lower anxiety at time a in some subjects, and lower at time b in some subjects.

I have this set up the following way:

subjects 293
Sessions 2
Session specific structurals
two conditions: Rest A = Session 1 and not Session 2
Rest B = Session 2 and not Session 1

one covariate: Change in anxiety between A and B (ChangeAnxiety)

I run the default preprocessing and denoising

in the second level analysis i select all subjects (0) and ChangeAnxiety (1) and between conditions (Rest A and Rest B) 1 -1

I did not find any significant results, which is ok, but when I run the analysis

All subjects = 1 and Between conditions (RestA and Rest B) ) 1 -1, which should basically be the test-retest, I find massive differences, (T-values >8 in the DMN, see attached screenshot)

Am I setting things up incorrectly when I specify two sessions, and two conditions?

Best
Clas

Dear Clas,

It seems to me that both contrasts are testing different things.

In the first case, it tests whether there is any connectivity change between the two sessions that can only be correlated with the change in anxiety, regressing out average connectivity of all subjects.

In the second contrast, it tests whether there is any connectivity change between the two sessions, whatever is the cause.

As I understand it, the interpretation would be that there are connectivity changes between the two sessions, but they cannot be explained by the change in anxiety.