Dear SPMers,
I have a question regarding the extraction of VOIs for a DCM analysis.
I have setup up a GLM design matrix with 9 concatenated sessions. To
account for session effects, I have also included 9 covariate
regressors, according to the following scheme:
SCAN COV1 COV2 COV9
scan1-sess1 1 0 ... 0
scan2-sess1 1 0 ... 0
...
scann-sess1 1 0 ... 0
scan1-sess2 0 1 ... 0
scan2-sess2 0 1 ... 0
...
scann-sess2 0 1 ... 0
...
scan1-sess9 0 0 ... 1
scan2-sess9 0 0 ... 1
...
scann-sess9 0 0 ... 1
If I include a Global Normalization scaling, the VOIs look like the
right eigenvariate in the attached figure. Using these VOIs for a DCM
analysis produces significant parameters as expected.
However, I would like NOT to do Global Normalization scaling (a reviewer
explicitely asked that we do not do so). If I do not scale for Global
normalization, the VOIs now look like the left eigenvariate in the
attached figure, with large spikes at between session transitions. The
DCM parameters obtained by using these VOIs are greatly reduced.
Is anybody aware of some methods to circumvent this problem (apart from
not concatenating sessions...)?
Thank you and best wishes,
Marco
--
Marco Tettamanti, Ph.D.
San Raffaele Scientific Institute
Facoltà di Psicologia
Via Olgettina 58
I-20132 Milano, Italy
Tel. ++39-02-26434888
Fax ++39-02-26434892
Email: [log in to unmask]
|