Hi SPMers,
I have a couple of practical questions concerning connectivity.
I am trying one model with SEM in order to compare the strength of
connections between two groups. Observed variables are the time course of 5
regions activated for all subjects at random level. In both groups this
specific model was validated in 7 out of 12 subjects (p > 0.05 for the chi
square value). May I validate this model for the whole group and go to the
second step (i.e. comparision of strenghts between groups).
I tried also to include all the subjects of one group in one single model
as in the figure 4 of Mechelli et al., Neuroimage 17, 1459-1469, (2002). I
would be then able to estimate all parameters without restrictions. But I
was dissapointed when I got an answer like "The model is probably
unidentified. In order to achieve identifiability, it will probably be
necessary to impose 120 additional constraints." I have to say that there
is already 1530 df !! Any ideas why it couldn't work here ?
Thank you in advance
Ps- I can not do DCM because my experience has a TR of 6 secs
----------------------------------------------------------------------------------------------------
Mario Aparicio
ULP - University of Strasbourg
-----------------------------------------------------------------------------------------------------
|