Hi Thais,

Design is fine. For the contrasts, only c1 and c2 are necessary. You already know that FA>0 (from the definition), so there is no need to make any test for the contrasts c3 through c8.

All the best,

Anderson


On 12 June 2017 at 18:35, Thais Minett <[log in to unmask]> wrote:
Dear FSL experts,

I would like to run 2 group analyses (with 3 demined covariables) on a DTI (tbss) data.

My matrix looks like this:

        0.000000e+00    1.000000e+00    3.852140e-01    9.537659e-01    -9.333656e+00
        1.000000e+00    0.000000e+00    3.852140e-01    -1.674370e+01   1.406663e+02
        0.000000e+00    1.000000e+00    3.852140e-01    1.299210e+01    4.066634e+01
        0.000000e+00    1.000000e+00    -6.147860e-01   5.766907e+00    -1.093337e+02
        1.000000e+00    0.000000e+00    3.852140e-01    -1.746923e+01   -1.093337e+02
        0.000000e+00    1.000000e+00    3.852140e-01    1.114678e+01    1.906663e+02

...and so on...

 My contrasts (among others) are:

         ev1 ev2 ev3 ev4 ev5 ev6
c1     1     -1    0     0     0     0
c2     -1     1    0     0     0     0
c3     1      0    0     0     0     0
c4     -1     0    0     0     0     0
c5     0      1    0     0     0     0
c6     0     -1    0     0     0     0
c7    0.5  0.5   0     0     0     0
c8   -0.5 -0.5  0     0     0     0

Am I correct to think that contrast 7 is testing if the overall FA mean (both groups together) is >0?

My c3 showed significant results so did my c5. But my c7 didn't... I was expecting that if g1 mean is >0 and g2 mean is >0 that my g1+g2 mean would also be >0. Where did I get wrong?

Regards,

Thais Minett