> I have a block design with 3 runs (condition A) and 3 levels
> in a run (condition B).
Do you mean that those 3 runs were analyzed separately?
> Now, I want to do an analysis across my 10 S's. It seems to
> me I need to get something like the IRF or some mean
> signal value to do this but then that would be different for
> every analysis. Where do I go next? I want to know if the B
> effect varies over A across S's.
I assume that you did deconvolve for individual subject analysis. For group analysis you can summerize each IRF with the sum (so-called area under the curve) or mean over the span of each IRF.
It looks like you need to run a 2-way within-subject ANOVA. And in AFNI subject is also counted as a factor, so you run 3-way ANOVA. You might have to use the Matlab package for this analysis:
http://afni.nimh.nih.gov/sscc/gangc/
Gang