Dear Norbert,
I'm forwarding my answer to the SPM list as well as to my colleagues
who might give you more detailed information. I think it'd be better
if for questions of general interest you use the SPM list from now on
as I'm sure your questions will be of interest to many other SPM
users.
2008/11/8 Norbert <[log in to unmask]>:
> Dear Vladimir,
>
> I have been playing around with the DCM tool for EEG. Using the
> DCMexample.mat data and DCM and some of my own data. In general I could do
> everything, however the main problem I have is with the neuronal model B
> matrix. In the example B matrix there are checked all the forward and
> backward connections except for the lateral ones, but also added are the
> left A1 to Left A1 and right A1 to right A1 connections - why??? The manual
> is not very explicit on this, jus saying: "The B matrix contains all gain
> modulations of connection strengths as set in the A-matrices". This would
> mean also the lateral ones but not this left right A1 to A1
> connections...but probably I misunderstand something?
If you mean connections on the diagonal this has to do with recent
work of Stefan Kiebel
http://www.fil.ion.ucl.ac.uk/spm/doc/biblio/Year/2007.complete.html#sjk_dcm_intrinsic
Briefly, checking these buttons makes it possible to test the
hypothesis that experimental effects can be accounted for by
modulation of intrinsic connections (connections between the 3
sub-populations) within an area. You can either do that or not - this
will constitute different models.
>
> Second question - below the matrices there are 3 additional options dipolar
> symmetry constraints, optimize source locations and lock-trial-specific
> effects - the manual does not mention them?
>
These are very new options. Symmetry constraints are described in a
recent paper:
http://www.fil.ion.ucl.ac.uk/spm/doc/biblio/Year/2008.complete.html#matthias_dcm_constraints
Locking effects means that you force all the modifications on
connections to be of the same direction and magnitude. If you enable
optimizing source locations you will allow DCM optimization to change
dipole locations. Usually it only changed the orientation so that the
sources will retain their meaning.
> My third question is in relation with the articles in NeuroImage - one By
> David and the other by Garrido on DCM for evoked responses. In the Garrido
> article 3 models were tested and in each some of the connections were
> allowed to change (broken lines) in the F model the 3 forward connections in
> B the 3 backward connections and in FB both. How is this done? My first
> assumption was e.g., for F model Jus having two models one with both
> connections fixed and one with just the backward connections? Sounds
> awkward, especially for the FB were you would have no connections - just the
> laterals and input and all fixed?
>
This is exactly the meaning of B-matrix. In all 3 models you have
forward and backward connections but by checking and unchecking the
corresponding entries in the B matrix you either allow them to change
or keep them fixed. The assumption is that a particular connection
belongs to one of the 3 types so you don't need 3 B-matrices but only
one where you say whether a connection is allowed to change. The type
of this connection is determined by the A-matrices. So for F-model you
click at the entries of B corresponding to forward connections, for
B-model - to backward and for FB - both.
Best,
Vladimir
> Best regards
>
> Norbert
>
> Norbert Jaušovec
> Filozofska fakulteta,
> Oddelek za psihologijo
> Univerza v Mariboru
> Koroška 160, 2000 Maribor
> Tel: +386 2 2293606
> GSM: +386 41 387834
> [log in to unmask]
>
>
>
>
|