DCAN-Labs/MarginalModelCIFTI: MarginalModelCifti constructs and tests marginal models for surface and volume data.

Longitudinal/complex modeling of imaging data produces many issues. First, models become increasingly complex, leading to the potential for unknown batch effects to affect results. Second, complex models cannot be easily permuted to control for false positives and negatives (see: Eklund, 2015 as to why that’s important). Third, complex models are time-consuming to perform, which limits the ability to perform exploratory/discovery analyses on imaging data. MarginalModelCIFTI implements Bryan Guillaume and Tom Nichols’ marginal modeling approach to overcome these limitations. Critically, such an analysis can draw group (but not subject) level inferences from datasets. Making it useful to explore in the context of large datasets (e.g. ABCD/fcon1000/HCP).

install.packages("remotes")
remotes::install_github("DCAN-Labs/MarginalModelCIFTI")