Oh, sorry. The censoring is certainly simpler. Would you recommend I still do it in two steps, a 'nodata' step with censoring to get the stimulus design matrix and then a processing step where I include baseline polynomial regressors and the output of the first step? Otherwise, including the polynomials when censoring removes the start of their timecourse and they are no longer mean-zero - though I'm not sure of the consequences of that.
Thanks.