Hi Nicola,
I had a similar issue recently with a repeated measures design and TBSS, and was getting a similar error message with regards to rank deficiency. A model is rank deficient if you have more than one set of EVs that represents all variance. I think you get the rank deficiency message because you have two EVs specifying subject identity (1 and 2) and four EVs specifying levels (3-6), and the two sets of EVs explain the same variance. The rank deficiency warning is only of concern if you want to test for differences between subject identity and levels - but from your contrast matrix, it seems like you are only interested in differences between levels. Therefore, you should be able to ignore any complaints about rank deficiency with regards to your t-tests.
However, I notice your .fts file specifying an f-test has only four numbers: 1/1/1/1. I think with your current design, this may need changing to: 0/0/1/1/1/1. At the moment it is probably trying to calculate the mean of EVs 1, 2 (subject identity) and 3, 4 (first two levels), rather than EVs 3-6 (all four levels).
For my own repeated measures design, instead of entering an EV for each subject, I actually used the "group" feature to specify each subject (boxes to the left of EVs in the Glm GUI). This means that randomise will permute within the appropriate exchangeability blocks, according to a repeated measures design. See the post here https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;ad203a63.1109 from Matthew Webster which helped me with this point for my repeated measures TBSS design.
I think you have only one group, composed of 2 subjects, with 4 separate time points? So if you use the "group" feature to specify subject identity, you would have a design similar to:
Group Level1 Level2 Level3 Level4
1 1 0 0 0
1 0 1 0 0
1 0 0 1 0
1 0 0 0 1
2 1 0 0 0
2 0 1 0 0
2 0 0 1 0
2 0 0 0 1
Contrasts e.g.:
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
1 -1 0 0
etc.
F-test e.g.:
1 1 1 1
Then you need to make sure that you specify the .grp file when using the randomise command, e.g.
randomise -i all_FA_skeletonised -o FA -d design.mat -t design.con -f design.fts -e design.grp -m mean_FA_skeleton_mask --T2
Hope that helps,
Charlotte
|