Dear SPM mailing list,
I am running DCM for ERP with SPM12, and ran into a problem when running analysis on parameters from the BMA.
I am interested in doing some statistical tests on parameters extracted from the A-matrix of the BMA (comparing parameters between two groups as well as doing regression analysis with the parameters). However, for example when I extract lateral connection parameters some parameter values for some participants are -4 and not for other participants. As I understand -4 is the equivalent of the connection being absent. This makes sense, as the models in the Occams window for these particular subjects do not have models with lateral connections, thus they would not contribute to the BMA.
My question is, what would be the right way to deal with these -4 values, when running statistics on them (comparing connections between my two groups as well as regression analysis)?
1. Should I keep the values as they are (meaning that some will have the value -4) and do the statistical analysis?
Or
2. Should I exponentiate all the parameters since they are estimated in the log-space and adjust the exp (-4) to 0?
However, I’m not sure putting it to zero is the same as saying the connection is not there?
Thanks in advance!
Roshini
|