Hi Dianne and Ken,
I completely agree with my esteemed colleague Ken! Adding a separate
regressor for each 'bad scan' will remove all of the variance associated
with 'bad scans', not just the average variance associated with bad scans
that will be removed by the procedure Dianne originally described. Of
course, as Ken states, if the variance between bad scans is relatively low,
then this difference in modeling strategies shouldn't amount to much.
Cheers,
Daniel
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
Behalf Of Ken Roberts
Sent: Monday, April 17, 2006 5:04 PM
To: [log in to unmask]
Subject: Re: [SPM] User Specified Regression: Clarification
Hello Dianne,
I feel moved to expand slightly upon the answer provided by my esteemed
colleague Dr. Weissman. (hi Daniel!!) Modelling bad scans with a
single vector
containing ones for the bad scans will remove variance rougly associated
with
the average of those bad scans. Any variance between the bad scans
will not be
removed. I imagine that for many types of bad volumes (such as
excessive bursts
of noise, or from scanner artifacts from bad k-space datapoints), this
will not
be sufficient.
To completely remove the influence of every bad scan from your model, you
will
have to input a separate column for _each_ bad scan containing all zeros and
one non-zero value (at the location of that bad scan).
Ken
----------------------------------------------------
Ken Roberts
Woldorff Laboratory
Center for Cognitive Neuroscience, Duke University
(919) 668-1334
----------------------------------------------------
|