Dear all,
I'm a student in a lab rotation and i'd like to use DCM for the analysis
of an experiment similar to the MMN example in the manual.
In this example standards and rares are compared by using the values 0.00
and 1.00 (manual p.311). I would have expected -1 and 1, but there seems
to be some other logic behind. Could you shed some light on that? Which
values would I choose if i wanted a linear parametric design with let's
say 4 different values of the parameter? -3 -1 1 3?
So far I'm also not quite sure if the model I chose is a good one. I can
compare variants of my model with the Bayesian Model Comparison, but it's
hard for me to judge whether even my best model is reasonably good. This
is hardly a lack of the software of course but more a lack in my
knowledge. I found the "free energy" DCM.F value which I guess could be
helpful, but I'm not sure how to interpret this value. And the BMS rates
the models with a relative log-evidence. Relative to what?
Thanks for your assistance and sorries for the newbie questions!
Matthias
|