> In the afni regession pdf file, it states that if the HRF regressor has
> a max amplitude of 1, then the beta coefficients will represent %
> signal change from the mean.
>
> when you designate each timepoint in a block as '1' in the input into
> waver, you end up with amplitude greater then one (even when
> using peak =1), because the timepoints within a block overlap, even
> though each block is separated in time.
>
> can the beta coefficients still be interpreted as % signal change?

Yes, but the % signal change is interpreted in terms of the impulse response function, not the HRF of a whole block.

> Does this interpretation depend on how you normalize your data?

No, it should not.

> In the same afni doc they normalize the timecourse to the
> min(200, a/b*100)... shouldn't it be min(200,(a/b)*100)?

I don't know where you got the formula, but you are probably correct.

> we usually normalize to ((a-b)/b)*100 does this make a difference?

What are a and b here? During pre-processing or after individual subject analysis?

> waver doesn't come close to obeying the peak =1 unless your
> input file is 1's and 0s, not 0's and 10's as outputed by sqwave....
> is there a way to change what value is used in sqwave?

Set BLOCK(d,p) with p =1 as your rtype in option "-stim_times k tname rtype" in 3dDeconvolve instead of running waver. This would only work if the duration of all blocks are the same. See more here:

http://afni.nimh.nih.gov/pub/dist/doc/misc/Decon/DeconSummer2004.html

Gang