Hi,

Firstly, your design can be compared to the example here (only difference is this is for 4 levels and you have 3)

http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_1-factor_4-levels__.281-way_between-subjects_ANOVA.29

I would not use the "group" column as you did, especially for randomise, but presumably you did not load the design.grp into your randomise call.

Otherwise, it isn't surprising that an F-test isn't significant when a t-test is. This can happen for weak effects, since an F-test is 2-sided and a t-test is one sided (hence p-values for 1-sided tests will be smaller).

Cheers,

Jeanette

On Fri, Apr 26, 2013 at 9:22 AM, Patricia Pires <[log in to unmask]> wrote:

Dear Mark and FSL experts,

I wanted to ensure that the construction of my Glm design is correct. I have 3 groups and i want to obtain F results (global differences) and t results (group contrasts differences).

I proceed as following in Glm (images attached). Is the Glm construction ok?

And then I ran randomise. I have obtained F results in FA, MD, AD and RD. There were results in t contrast for FA and RD. However, although F results were evident in MD and AD (at --thresh=0.95), i didn't find any t contrast difference (at --thresh=0.95 as there were in F analysis). Only at --thresh=0.90 or less.

Could you please tell me if i proceed correctly?

Thanks in advance,

Patricia.