FBA stats clarification/interpretation

1

Dear MRtrix Users,

I have completed FBA and I need some suggestions and clarifications. I have the design.txt as follows:
group1 group2 age sex. I used two different contrast [1 -1 0 0] for group1>group2 for group difference by controlling age and sex; [1 -1 1 1] for group1>group2 for group difference by including age and sex effect. Are they right? If they are, what exactly the difference between the two conditions?

Also, I have reading and intelligence scores and I added them into my design matrix. My new design matrix is group1 group2 group1ReadingScore group2ReadingScore age sex. To have differences between two groups in their associations I used two different contrast [0 0 1 -1 1 1] and [0 0 1 -1 0 0]. Are they fine? Again, if both are fine, what would be the best interpretation?

Thanks,
Cmid

2

Hi @cmid,

I think there’s a fundamental misunderstanding here about exactly what role the contrast matrix plays in the formulation of a hypothesis using the GLM. I’ve not written a documentatin page on the GLM, and frankly doubt the likelihood of ever finding the time, but there are resources out there by those with more experience with this stuff than I have.

For instance, what is your intended difference between:

… group difference by controlling age and sex …

and:

… group difference by including age and sex effect …

?

What’s actually happening when you construct your contrast as:
[1 -1 1 1]
, is that your hypothesis becomes (using the FDC metric as the example):

      d_FDC      d_FDC     d_FDC    d_FDC
H_1:  -----   -  -----   + -----  + -----  > 0
     d_Group1   d_Group2   d_Age    d_Sex

If this is not your intended hypothesis, then the GLM construction is incorrect.

Regards
Rob