Dear Experts
I have run a cmm_NMDA DCM using LFPs for a series of subjects. Within
these subjects there are two populations and two groups; i.e.
pop1-group1
pop2-group1
pop1-group2
pop2-group2
My intention is to find an optimally reduced model using PEB and to
compare model parameters across pops and groups, but I am unsure if I am
executing this correctly, and if so, how to interpret the results.
I have done this in a two-stage process: a PEB for the main effects and
a PEB for second level effects across groups:
1) The DCMs are first inverted for the full model for each subject,
loading them into a cell array (GCM). Then I have built a design matrix
(M1.X) containing 0 & 1's for the PEB that has a column for each
pop-group combo (with a first column entirely of ones). Then I run it as
follows:
PEB1 = spm_dcm_peb(GCM,M1,fields);
BMA1 = spm_dcm_peb_bmc(PEB1);
These above results I believe I can interpret. However...
2) To compare across groups I have created a design matrix (M2.X)
containing -1, 1 & 0's for each desired contrast - in this case I have 4
contrasts (5 columns in total due to the first column of ones). This I
ran using the following (please note the change in script used):
[BMC2,PEB2] = spm_dcm_bmc_peb(GCM,M2,fields);
BMA2 = spm_dcm_peb_bmc(PEB2)
QUESTIONS:
A) Are these the correct script choices for this scenario?
B) How do I interpret the result in BMC2? It speaks of 8 models, one of
which wins based on it's free energy, yet I do not understand what these
8 models actually are... The help possibly implies the winning model
would be given in PEB2.M, but there is no .M.
C) As I have run the PEB twice (once for within-groups and once for
between-groups) I wonder if these results cannot be legitimately
consolidated.. Any thoughts?
Any help appreciated,
Thanks,
Natalie
|