You can use FEAT. For the 1st level, disable the temporal filters, use a simple regressor coding your independent variable, don't convolve that with any HRF, and adjust the registration options so that it's appropriate to your kind of data.

Another option is to use PALM. However, hardly compound symmetry will hold. You can still use whole-block sign-flipping (one block per subject), and permute subjects that have the same number of observations.

A third option is Bryan Guillaume's SwE toolbox: https://www2.warwick.ac.uk/fac/sci/statistics/staff/academic-research/nichols/software/swe

All the best,

Anderson

On Tuesday, 12 September 2017, Fulvio Da Re <[log in to unmask]> wrote:

Hi all,

I'd like to start a structural MRI longitudinal analysis in my Alzheimer's Disease cohort, studying the atrophy progression. In particular, I'd like to fit a linear mixed effects model using fixed effects of disease duration and a random intercept term for patient.

Most of my subjects have more than 2 timepoints (actually, they have a variable amount of timepoints between 2 and 6), so I don't think SIENA will fit my requirements. Neither FEAT should do that, since it was designed for fMRI.

Any suggestions?

Thanks so much for helping

Fulvio Da Re

PhD student

FTD Center, University of Pennsylvania