Hello FSL experts,
I have read a handful of posts and the FSL GLM wiki for guidance, but I would greatly appreciate some confirmation as to how I am designing my GLM model and clarification on what the f-tests are showing.
I have a 3 group design with controls (NC), MCI, and Alzheimer's (AD) subjects. We have one continuous cognitive variable of interest (cog) and two continuous covariates Age and Education (Edu).
Our question is whether there are differences in white matter FA both between and within our three groups that is dependent on our cognitive variable while holding constant any effects of age and education.
We started by using t-tests to show that we have FA differences between groups while regressing effects of (centered) age and education...
EVs
NC MCI AD Age Edu
1 0 0 5 1
1 0 0 -5 -2
0 1 0 3 2
0 1 0 -4 -1
0 0 1 4 2
0 0 1 -3 -2
Contrasts
NC MCI AD Age Edu
NC>MCI 1 -1 0 0 0
NC>AD 1 0 -1 0 0
MCI>AD 0 1 -1 0 0
MCI>NC -1 1 0 0 0
AD>NC -1 0 1 0 0
AD>MCI 0 -1 1 0 0
However, we did not run the accompanying f-test to show overall differences between the three groups. I have seen conflicting posts on how to set this up. Do I click the "F1" box for contracts 1 and 2 only, for contrasts 1, 2, and 3, or do I need to create a new contrast model, like this one below, and select the F's for contrast 1, 2, and 3?
NC MCI AD Age Edu
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
Which one of these options is correct and what will the output ftest.nii show?
Next we ran a model to see how our cognitive variable is related to FA in our sample as a whole...
EVs
cog
21.4
15.6
6.9
5.4
-12.2
-18.7
Contrasts
cog
pos_cog 1
neg_cog -1
We then ran a model to test for our interaction between cognition (centered across groups) and patient group as follows...
EVs
NC MCI AD cogNC cogMCI cogAD Age Edu
1 0 0 21.4 0 0 5 1
1 0 0 15.6 0 0 -5 -2
0 1 0 0 6.9 0 3 2
0 1 0 0 5.4 0 -4 -1
0 0 1 0 0 -12.2 4 2
0 0 1 0 0 -18.7 -3 -2
Contrasts
NC MCI AD cogNC cogMCI cogAD Age Edu
cogNC>cogMCI 0 0 0 1 -1 0 0 0
cogNC>cogAD 0 0 0 1 0 -1 0 0
cogMCI>cogAD 0 0 0 0 1 -1 0 0
cogMCI>cogNC 0 0 0 -1 1 0 0 0
cogAD>cogMCI 0 0 0 0 -1 1 0 0
cogAD>cogNC 0 0 0 -1 0 1 0 0
First of all, does this model appear to be set up correctly?
Is there an additional f-test I need to run before I am allowed to make these between group comparisons with cognition? If so, would it just be the effect of cognition on the groups with F's selected in the same way as the first model (except now the groups are cogNC, cogMCI, and cogAD).
Finally we want to see the effects of cognition within each of our groups. To do that we used the following model...
EVs
NC MCI AD cogNC cogMCI cogAD Age Edu
1 0 0 21.4 0 0 5 1
1 0 0 15.6 0 0 -5 -2
0 1 0 0 6.9 0 3 2
0 1 0 0 5.4 0 -4 -1
0 0 1 0 0 -12.2 4 2
0 0 1 0 0 -18.7 -3 -2
Contrasts
NC MCI AD cogNC cogMCI cogAD Age Edu
pos_cogNC 0 0 0 1 0 0 0 0
ned_cogNC 0 0 0 -1 0 0 0 0
pos_cogMCI 0 0 0 0 1 0 0 0
neg_cogMCI 0 0 0 0 -1 0 0 0
pos_cogAD 0 0 0 0 0 1 0 0
neg_cogAD 0 0 0 0 0 -1 0 0
Again, is this model accurately testing my question, and are there prerequisite f-tests that I need to show first to avoid a reviewer's wrath.
Thank you for any help you can provide,
Anthony
|