Dear Art,
Hello. I cannot point you to the lines of code where the parametric
regressors are created but here is what I do know. If you have a
condition which alternates between 'on' and 'off' one regressor will
capture this mean effect of the task. If you believe that the
response systematically varies around this mean then a parametric
modulator can capture this variance. So the parametric modulator is
orthogonal to the mean regressor (the one where each trial/block has
the same amplitude). One good example I have tested out is a task
where I have three memory load levels. These three levels can be
modeled with three regressors, then a linear increase as a function of
load can be tested with a contrast. OR the model can be done with a
single regressor coding every trial/epoch regardless of the load
level. Then a second regressor (the parametric modulator) can be added
to the model that captures changes in the amplitude of the BOLD signal
as a function of load level. The results from the two models give the
same results but the parametric modulator has one more degree of
freedom.
So the stats are not "special" but fall within in the context of standard GLM.
I hope this helps,
Jason
On Thu, Mar 3, 2011 at 1:36 PM, Stephen J. Fromm <[log in to unmask]> wrote:
> I don't know, but the obvious way to go about this on a Linux/Unix/Mac box would be to cd into the code directory and type
> grep -i modulation *.m
> As for comparing versions, use the Unix "diff" utility.
>
--
Jason Steffener, Ph.D.
Department of Neurology
Columbia University
http://www.cogneurosci.org/steffener.html
|