Autoregressive models for time-dependent modalities

[arokem]

One thing that occurred to me in learning about the PET b-spline model is that the time-dependent modalities (i.e., PET and fMRI) may benefit from some form of multi-variate auto-regressive modeling approaches. This would allow the model to capitalize on correlations between different regions, etc. Not sure if it makes sense at all, and certainly not something we need to do anytime soon. But thought I'd record the thought.