help > GTG for event-related design
Showing 1-2 of 2 posts
Display:
Results per page:
Oct 6, 2016  05:10 PM | Alex Rainer
GTG for event-related design
Dear Jeffrey,

Thank you for sharing the toolbox. Could you please give us your thoughts on which features of an event-related design would be important for GTG (e.g., the number of trials, the length of the trials, the length of the event(s) that GTG analysis is applied to, etc.)? I see from the manual that GTG is primarily for block design. For event-related design, it uses beta series though this has not been tested vigorously. Is this still the case?

Best,

Alex
Oct 6, 2016  11:10 PM | Jeffrey Spielberg
RE: GTG for event-related design
Hi Alex,

Off the top of my head, I'd say that you want to maximize the # of trials and the separability of the trials (e.g., via ISI jittering, stimulus order), rather trial length, etc.  Basically, you want to maximize signal detection for the effects of interest (rather than HRF estimation).  

As for GTG's implementation of BSC, we've tested this out with a few different data sets, but have not yet published anything.  In other words, I'm fairly confident that it does what it's supposed to, but want people to know that this aspect of toolbox has not been tested nearly as extensively as the rest.  Also, you should definitely avoid using the "node-specific mean HRF" at this point, as we have not evaluated whether this is any better than (or even comparable to) using an assumed HRF.  

Best,
Jeff