Testing repeated measures using new fixelcfestats

1

Hello,

I would like to use FBA to perform a repeated measures test between two scans for the same subjects. I have two main concerns: 1) creating the proper design matrix (and contrast file), 2) using the appropriate fixelcfestats options.

For the design matrix, going by what’s described in FSL’s GLM documentation, I’m assuming this design would be appropriate? (assuming I only have 5 subjects):

1 1 0 0 0 0
1 0 1 0 0 0
1 0 0 1 0 0
1 0 0 0 1 0
1 0 0 0 0 1
-1 1 0 0 0 0
-1 0 1 0 0 0
-1 0 0 1 0 0
-1 0 0 0 1 0
-1 0 0 0 0 1

And contrast matrix

1 0 0 0 0 0

Is that accurate? If not, how would you suggest I design my matrix?

For using fixelcfestats, I went through the new documentation and I’m a bit overwhelmed by all the new options. I’m not sure if I need to add any extra options for my setup to work or just use fixelcfestats directly. Are there any specific options you think I should use?

I guess my questions are directed at @rsmith, but suggestions from anyone are more than welcome.

I hope you’re all staying safe and well.

Best,
Joe

1 Like
2

Hi Joe,

While that design will work as presented, it has the limitation that you are unable to incorporate any sort of per-subject nuisance regressor into your model, as they would be wholly collinear with the per-subject means.

If you follow your link above, but scroll down beyond the “FEAT” details to “randomize”, there it describes the preferential approach: explicitly compute the difference between the two scans for each subject, and then perform a one-sample t-test (-errors ise in fixelcfestats). Anderson Winkler’s 2014 manuscript is the master document for such things.

I’m a bit overwhelmed by all the new options

Imagine how I feel :sweat_smile:

Cheers
Rob

1 Like