Hi Giuseppe,

I believe that Doug suggested to use the maximum value
of the convolved stimulus function in response to someone
comparing regressors of different types. One was based on
a stimulus function, passed through waver, and two were
based on speech movement information.

Since the movement data was already in a curve form, not
just a binary stimulus form, in order to compare percent
signal changes between the movement data and the stimulus
response data, they needed to be normalized. So they were
normalized based on the height of the curves, after any
convolution.

At any rate, I do not believe Doug was suggesting that as
a standard.

His general suggestion for comparing the percent signal
change from stimulus types is to normalize the actual run
data (maybe EPI) to an average of 100, and can be found
here:

[afni.nimh.nih.gov]

This technique will be used in the next HowTo, where
percent signal change will be discussed in fair detail.

- rick