I’m trying to apply MVPA using PyMVPA (http://www.pymvpa.org/tutorial_eventrelated.html) on a dataset that I have, but have a problem with it. With the following function, we can fit a model of the hemodynamic response to all relevant stimulus conditions and later feed our classifier with the results of hrf modeling rather than raw BOLD data; time_attr is an attribute with a time stamp for each sample. Also, (‘targets’, ‘chunks’) indicates that we want a separate model for each stimulation condition (targets) for each run of our example dataset (chunks):
evds = fit_event_hrf_model(ds,
… condition_attr=(‘targets’, ‘chunks’))
As I understand from this function, it still gets samples (raw BOLD volumes) as inputs and needs us to attribute time stamps to each sample. So what if I have jittered my data and events don’t start at the beginning of volumes? In this case, labeling samples isn’t really reasonable because one sample can contain data of two different conditions. I hope my question is clear. I’d appreciate it if anyone could help with this!