I am trying to understand the DCM results from my analysis on 20
subjects. The analysis is done in SPM5. The experiment has 3 conditions
P N and neu. They are fed as inputs c to the Area F (sphere of 4 mm
containing 15 - 33 voxels). We made 2 models - both containing 2 areas F
and G with matrix a (effective connectivity) containing both directions
(from F to G and from G to F). First model has only 1 modulatory effect
- condition N. Second model has 2 modulatory effects P and N.
The question we have is : In 3 cases in model2 and 1 case in model1 the
probabilities pA are NaN. For the same areas the alternative model for
the same subjects (and areas) has finite pA's and pB's. What does this mean?
How can I treat those models?
The idea is to compare two models and see if the condition P has
modulatory effect on the connections and what effect. Therefore we are
trying to compare these 2 models. But when pA's are NaN there is no way
to compare the models.
I would really appreciate if somebody could answer me these questions.
P.S. I checked the models thoroughly, and I made no mistakes. The
previous emails from SPM list discuss the NaN problems in SPM2 and there
is one explanation for the errors made while creating the models in
SPM5. But I didn't make any errors.
Branislava Curcic-Blake, PhD
Cognitive Neuropsychiatry group
BCN Neuroimaging Center (NIC)
University Medical Center Groningen
Antonius Deusinglaan 2
9713 AW Groningen
[log in to unmask]