Gang Wrote:
-------------------------------------------------------
> > More specifically, a parameter estimate
> representing the joint effect of Stimulus1
> > and Stimulus2, excluding the other_effects.
>
> Unfortunately such a test can't be represented by
> one effect estimate, but two separate ones: beta1
> and beta2.


I see.

Just to follow up on an earlier part of the thread:

>>>My reservation about the first test in my previous response is that it weighs the effects of the two stimuli equally and one effect has the opposite sign to >>>the other, which is rarely the case in the reality of FMRI, except for some special scenarios. More specifically the first test compares the following two>>> >>>models:

>>>Full model: y = Effect_Of_Stimulus1 + Effect_Of_Stimulus2 + other_effects + residuals

>>>Reduced model: y = Effect_Of_Stimulus1 + Effect_Of_Stimulus2 + other_effects + residuals, with the constraint of beta1 + beta2 = 0


Well, what if Simulus1 and Stimulus2 represent time courses from, let's say, the default mode network for resting state analysis. With this way, we would be able to obtain and R^2. We could then get the correlation coefficient and we could identify the direction of correlation coefficient based on the sign coefficient and t-value of the joint correlation. Again, this is all done at the individual subject level. The thinking is that, though not identical, both seed time courses should have overlapping activation maps if we were to run seed-based correlation separately for each time course. Are we losing anything with the gltsym approach above? Or should we try a different approach?

Michael