 |
 |
Excellent point - FIR is way more flexible .. but as you know comes with it's set of problems too. To get back to the point of using derivatives - the variable duration model works well for detection but not so much when comparing multiple conditions against each other. I also agree (but this is mentioned in Grinband's paper) that other areas where there is no modulation will be less well modeled. So far (but I have to admit Jeanette and I have done limited (independent) simulation work on it) taking the mean over all trials for duration (do note I say all trials across all conditions - not condition specific) to model each regressor and add modulation by RT specific to each regressor seems to work best (as in control better your type 1 error / power) in the case you want to compare/contrasts multiple conditions -- hope this clarify a little the context Best Cyril ________________________________ From: MRI More <[log in to unmask]> Sent: 15 October 2016 13:48:39 To: [log in to unmask]; PERNET Cyril Subject: Re: the calculation of the latency of the BOLD response and the explaination of the temporal derivative of parametric regressor > using fixed mean duration (across all trials) + parametric regressors to model changes works better ... Works better in which regard / under which assumptions? I think it's important to provide the exact settings to avoid overgeneralizations. Predictors based on condition mean RTs can e.g. be problematic in case the conditions evoke RT-independent activations. Due to the differently scaled predictors the estimates would be scaled differently as well and possibly lead to a condition difference despite the two conditions resulting in signal changes of the same amplitude and the same temporal course. Given the massive amount of fMRI studies, often relying on strong assumptions for the predictors (e.g. the simple "on"/"off" pattern, the canonical HRF), it is quite funny that evidence for the predictors and the concepts behind is actually still rather anecdotal. Which would speak for flexible sets of predictors like the FIR approach, which are impractical for many experiments though. Then again, one could argue that one shouldn't conduct any such experiments at all instead of turning to "somewhat" more flexible sets with still rather strong assumptions (e.g. the "on"/"off" behavior, a single peak). Best Helmut