I got strange results when I compare 30 DCMs by
spm_dcm_compare.m in SPM8.
($Id: spm_dcm_compare.m 3363 2009-09-04 15:11:19Z christophe $)
The model 1-15 have different connections among 5 regions, and contextual
modulation is allowed for all connections including recursive (i.e.,
regionA --> regionA) ones.
The model 16-30 were the same as the model 1-15, but the recursive ones
When I compared model 1-15, the model 13 won.
When I compared model 16-30, the model 28 won
The model 13 and 28 are the same except the recursive modulation.
Surprisingly, when I compared the model 1-30, then model 15 won.
In the comparison of 13, 15, and 28, the model 15 won.
I explored several combinations e.g.
model 1-15--> model 15
model 5-15 --> model 15
model 2-15--> model 15
mode 1-13 and 15--> model 13
model 1-14, 16-30--> model 13
I wonder why this happens, and which is the best model.
The model 15 is the full connection model, i.e., all possible
connections are modeled including
the recursive ones
Thank you in advance for your kind explanations!
Max Planck Institute for Human Cognitive and Brain Sciences
Stephanstrasse 1a, 04103 Leipzig, Germany
PO BOX 500 355, 04303 Leipzig
E-MAIL [log in to unmask]