Fixelcfestats: specify exchangability-block information

Hi Dave,
Looking to try Orjan’s code, but I’m not willing to rely on anything that has not been unit-tested. Do you have unit tests for your permutation testing somewhere? If so, I’ll try to modify it for this purpose.

If not, how would you see this code being checked? Other than checking that the distributions are correctly generated, perhaps some normal data could be used, and results compared against a one sample t-test?

No sorry. Permutation testing is reasonably trivial to implement, so unit tests for this have not been high on my priority list. So far I’ve only checked the GLM ttest output against other stats packages. The stats code in MRtrix is going to get more advanced in the near future (with some additions from @rsmith), so I plan on writing some tests then.

Dear Dave,

do you have any progress with exchangability-block option for fixelcfestats? I would be eager to use fixel-based analysis in my study, however I would need not only one-sample t-test, but 2-way ANOVA (ideally with some additional covariate): I have 2 groups, each subject measured twice and I am interested in differences between groups in temporary change of WM mictrostructure.

Regards,
Antonin Skoch

1 Like

Hi Antonin,
Yes the next release (to be released soon in the new year) will enable users to specify the permutations manually (using a text file that can be generated with FSL Palm, where you can specify the EB).
Cheers,
Dave