HI Everyone - My question is really and of topic, sorry
I’m analyzing dwMRI data. In particular, I would like to compare mean diffusivity (MD) scalar maps of participants before and after a motor task.
The research design comprises two sets of three MD maps, and I would like to test for changes beween task and control and changes across time in diffusion values
To test my hypothesis, I would like to structure a mixed ANOVA with two conditions (between participants, task and controll) and three levels (within sets A, B and C).
How would you structure the design matrix? Can NILearn run this kind of analysis? Do I need to use fsl_glm or SPM for that?
I can’t comment on how NiLearn deals with such things. The MRtrix3 statistical inference commands operate on the same GLM backbone as FSL, and so a GLM constructed for FSL would likely be compatible with MRtrix3, with the possible caveat that with MRtrix3 a global intercept column is not added automatically, it must be added explicitly if desired (to facilitate experimental designs where it is desired to exclude such).
I can’t decipher your description to the point of being able to express it in matrix form. You state:
two sets of three MD maps
, but I don’t know along which experimental variable there are two measurements, and along which experimental variable there are three measurements. In particular you state:
within sets A, B and C
, which contradicts the prior “… two sets …”.
Tests “between task and control” and “changes across time” would conventionally in GLM-land be interrogated as two distinct hypotheses, potentially with differences in how data are prepared for the model. But I’m not sure from your description whether you are trying to do something more advanced than this.
I’ve often remarked that one’s understanding of their experimental design and hypotheses needs to be adequate to facilitate their translation into numerical form. But if one is not familiar with the GLM, a similar logic holds, in that it needs to be feasible for someone else to translate your description of your experimental design and hypotheses into numerical form. I can’t quote a design matrix for someone to copy and paste unless I have adequate confidence that my understanding of their use case is correct, and that their use of such is not going to be in one of myriad ways silently erroneous.
