hi Thomas, please allow me to suggest and promote our work on this topic: https://www.ncbi.nlm.nih.gov/pubmed/25304775 pdf : https://hal.inria.fr/file/index/docid/974190/filename/paper.pdf it comes also with some code : https://github.com/fabianp/hrf_estimation Sorry for the self promotion, but I think it can be relevant for your work. Best, Alex -- Alexandre Gramfort, PhD Inria, Université Paris-Saclay http://alexandre.gramfort.net [log in to unmask] On Fri, Mar 16, 2018 at 5:17 PM, Thomas Cope <[log in to unmask]> wrote: > Dear SPM community, > > > Is there any way to apply higher order autoregressive models than > AR(1) during classical first level model estimation? > > > I am attempting to try out the methods described here: > http://ieeexplore.ieee.org/document/7552348/ for better deconvolution > of my MVPA conditions in a rapid event-related design. The authors tell me > that they implemented their higher order AR models using Bayesian > estimation. I have tried this but, unfortunately, for my complex design > and high resolution (1.4mm isotropic whole brain) data this seems > computationally unfeasible - our high performance cluster has been working > on a single subject for a week and is only half way through. > > > Can anyone point me in the direction of a way to either pre-process my > data before estimation or to hack the SPM files to allow higher order AR > without the need for Bayesian estimation? > > > Best wishes, > > > Thomas > > >