I have a jittered event-related fMRI data. I need to do MVPA analysis, but I am confused about which GLM approach to perform. As I know, there are two common approaches LSA (Least Squares - All) and LSS (Least Squares - Separate). So to choose the best for my data, I should estimate the trial-to-trial variability and scan-to-scan (scan noise) variability, and incoherence of both. I know to papers focused on trail-specific response (Mumford et al. (2012), H Abdulrahman et al. (2016). They mentioned the trial and scan variability. Unfortunately, I failed to understand how to estimate them.
Any suggestion will be appreciated.