Hey everyone, I have 2 questions on glmsingle.
Can it handle multiple durations ? For example, if I have both a block and an event in my design.
Also, do I need to provide nuisance regressors, like motion outliers ?
1 Like
Hi @Paradeisios,
I don’r see why it wouldn’t, but in general I caution against directly comparing block vs. event related design, as block designs almost always achieve higher effect sizes (at the expense of ecological validity).
From the documentation:
- Projecting out nuisance components – Pipelines often remove/project-out nuisance components (e.g. motion regressors, ICA derived noise, low-rank approaches like NORDIC, etc.) from fMRI time series in the course of pre-processing. While you can do this, and GLMsingle will likely still work, this is not quite recommended. This is because such pre-filtering approaches risk bias. And, GLMsingle’s approach is to attempt to learn these types of nuisance components from the data themselves, so it is a bit ironic to use both approaches simultaneously. Moreover, it is possible that using a pre-filtering approach will actually cause strange complications with GLMsingle, so beware.
Best,
Steven
I think this is covered in this part of the doc:
1 Like