Run a t-test comparing the HRF parameters only,
i.e. contrast 1 0 0 -1 0 0
The derivative and dispersion allow for a better model fit, but the HRF parameter is generally what is of interest (the amplitude of the BOLD changes). The f-test you describe allows for the detection of differences in the shape of the HRF in a non-specific way. in some cases this may be interested, but I suspect what you really want is what comes out of the t-test.
Best of luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
www.neuranalysis.com
[log in to unmask]
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of Marco Caviezel
Sent: May-15-17 10:15 AM
To: [log in to unmask]
Subject: [SPM] Time and dispersion derivates in T- and F-Contrasts in 1st level GLM
Dear SPM experts,
I have a question addressing time and dispersion derivates in T- and F-Contrasts in the 1st level analysis.
Assuming, having the following GLM:
FamousFaces(HRF); FamousFaces(TimeDerivate); FamousFaces(DispersionDerivates); NonFamousFaces(HRF); NonFamousFaces(TimeDerivate); NonFamousFaces(DispersionDerivates); 6xMotionParameters
So far I understood that if I include the time and dispersion derivates in my GLM, i have to use an F-test (e.g. [eye(3),eye(3)*-1] for differences between FamousFaces and NonFamousFaces), but these F-tests are not directional.
For a directional statement (e.g. more “activated” in FamousFaces than NonFamousFaces), I would use a T-test in the same GLM (e.g. 1 0 0 -1 0 0).
My question now is if this is correct, because I'm assuming that in this contrast, the time and dispersion derivates act as nuisance parameters, which are “correcting” for time and dispersion of the HRF.
Is this assumption correct?
So, would it be better to do a new GLM without time and dispersion derivates, or is there a method to combine the HRF, the time derivate and the dispersion derivate and perform a directional test?
Best, Marco
Marco Caviezel
PhD Student
University of Basel
|