I believe you are correct, as per my understanding.
I think "regressing out" is in fact correct here, however you may be more comfortable with a term such as "accounting for" or whatnot. The PMs are in fact part of a regression, which is what SPM does (the GLM is a form of least squares regression).
The reason the PMs matter is because of the use of SPm_ortho, which orthogonalized them in the design matrix. If they were not orthogonalized (which changes the regressors), order wouldn't matter.
Best,
Colin
-----Original Message-----
From: Mike [mailto:[log in to unmask]]
Sent: September-09-16 5:41 PM
To: [log in to unmask]; Colin Hawco
Subject: Re: Two parametric modulator
Dear Colin,
Thank you. So, if I have two parametric modulators (pm) A and B and I want to know (1) brain activity modulated by A while regressing out the effect of B, and (2) brain activity modulated by B while regressing out the effect of A, I have to construct two first level models, with the first using B as the first pm and A the second for (1), and another one with reverse order for (2), right?
Also, I wonder about the term "regressing out." This is usually use in multiple linear regression, when we have more than two covariates, and their order doesn't matter. I'm not sure if this term is appropriate for parametric modulation in which the order matters.
Mike
|