Dear SPM group and Dr Zeidman,
I am looking to use DCM PEB comparing two groups with longitudinal data. In brief, it is a patient vs control study, with three time points of task fMRI, where the effects of interest are how the effective connectivity and task behaviour of A condition > B condition change over time. However, the longitudinal data has attrition; 59 subjects at baseline (30 patients vs 29 controls), 38 subjects at time point 2 (19 patients vs 19 controls) and 29 subjects at time point 3 (12 patients vs 17 controls).
The most relevant approach I could find was the multi-level PEB, demonstrated in Park et al., 2017 (https://www.frontiersin.org/articles/10.3389/fneur.2017.00346/full) - however, I am uncertain how to handle the missing subject data. Is this a situation where you would need to specify priors on subjects with missing data when estimating DCMs at the first level?
Regards,
Bjorn
|