Dear Matt & Roar,
I think what you primarily describe would be a paradigm example for
non-inferential, data-driven exploratory data analyses as opposed to the
model-led inferential data analysis routinely employed by SPM.
Implementation of such tools are the independent component analysis offered
by MELODIC of the FMRIB Software Lib or commercially by the BrainVoyager and
the temporal fuzzy cluster anaylsis offered by Evident, just to name a few.
However, there are other way to define and tackle the problem within SPM but
I will leave that and listen to the experts here.
Best regards-
Andreas
****************************************************************************
******************************
Dr. Andreas J. Bartsch phone: +49 (0)931-201-0
Division of Neuroradiology,
ecr.: -5791
BJMU Wuerzburg pager:
#5325
Josef-Schneider-Str. 11 fax: +49 (0)
931-201-2685
97080 Wuerzburg email:
[log in to unmask]
Germany
[log in to unmask]
****************************************************************************
******************************
----- Original Message -----
From: "Matt and Roar" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, December 28, 2001 3:19 PM
Subject: With and without high pass filtering?
Dear SPM users,
We'd like to post this slightly unusual question again about high pass
filtering and signal drift. We are
investigating subtle changes in functional anatomy in the resting wake
state
across a period of ~20mins. We have specific hypotheses regarding the
regions of interest, and hope to go on to probe other states of
consciousness such as sleep. Because we do not have a classical "on" -
"off"
cognitive paradigm, it is hard to probe confounding factors, specifically
signal drift. Furthermore, the types of activity changes we predict could
very well be removed by using high pass filtering.Thus, we write to inquire
about ways to tease apart true signal changes from signal drift, even
changes such as the ones we expect in this paradigm, while still being able
to feel confident our results. Is there a safe way of analyzing such data
without high pass filtering, and if so what are the inherent dangers. If
not, is there a control task/study we could implement so as to make sure
such changes are not due to drift.
Thanks in advance for your help on this :)
Much appreciated.
--Matt Walker and Roar Fosse
[log in to unmask]
|