Hi FLS users,
I am having trouble setting up a model for Dual Regression analysis. My goal is to find brain regions within a network of interest (dorsal visual network) that covary with the scores to a questionnaire.
I have three groups (AN, BDD, CON), the scores to one questionnaire (EDE) only for the AN group.
I have age across all 3 groups, but EDE for only one group. Age is a nuisance regressor, and EDE is a covariate of interest.
My DVN (dorsal visual network) is a component extracted from a group ICA, which was run including subjects from all the groups.
This component was identified as dorsal visual network by using fslcc to compare components to Smith networks.
I have two options that I am considering in my head, and am not wild about neither of them:
First option:
to set up my model including all the groups (including AN, BDD, and CON). See the models that follow.
EVs
AN BDD CON age EDE
1 0 0 25 4.48
1 0 0 20 3.25
1 0 0 24 1.59
0 1 0 27 0
0 1 0 25 0
0 1 0 19 0
0 0 1 28 0
0 0 1 25 0
0 0 1 33 0
0 0 1 20 0
Contrast & Ftest
EV1 EV2 EV3 EV4 EV5
C1 1 0 0 0 0
C2 0 0 0 0 1
For results of regions covarying with EDE, I would look at the stage 3 results (tfce_corrp_tstat2) for the ic corresponding to the DVN component identified from the group melodic across all 3 groups.
And, since the DVN was extracted from all subjects I thought it might be more sound to include all the groups.
am I setting up Contrast2 correctly?
Does this identify regions that covary with EDE for only the AN group, since others have 0 value?
I could use some suggestions.
Would these regions represent areas where EDE covary for all the groups, including the un-wanted groups?
Would I be including a fake effect of EDE scores equal to 0 for BDD and CON?
Second option:
to set up a model including only the AN group. See following models.
But it seems that this would require a new group melodic with only the AN group, then a new DVN identified using fslcc, - and thus a different DVN component that applies only to the AN group.
EVs
AN AGE EDE
1 25 4.48
1 20 3.25
1 24 1.59
Contrast & Ftest
EV1 EV2 EV3
C1 1 0 0
C2 0 0 1
We want to identify regions that covary with EDE within the DVN
Is there a better way to do this?
Thanks for your help.
|