Dear Mina,

The residuals are stored in DCM.R.  You could use these to compute classical goodness-of-fit (GOF) measures like R2.  Note, however, that there is no such thing as "absolute" goodness of fit:  any classical GOF like R2 is just another model comparison.  This is not obvious because the reference model is implicit and extremely simple (a model consisting merely of a constant [intercept]).  See any good stats textbook for details.

Best wishes,

----- Urspr√ľngliche Mail ----
Von: Mina Khoshnejad <[log in to unmask]>
An: [log in to unmask]
Gesendet: Freitag, den 25. Juli 2008, 06:56:55 Uhr
Betreff: [SPM] How to look at residuals in the DCM output



I want to look at how well a DCM model can generate the time courses of the regions.

Although with Bayes Factors for each model, you can tell which model is the best, It does not tell you how good it emulate the time courses of the regions, for ex: you can have a poor correlation between the estimated and real time courses of the regions for the best selected model. would this mean that the hypothesis is wrong?


My question is what is a measure of a good fit here? or to what degree the estimated and real time courses should be similar? The vidence term should represent it some how.


Does anybody know if the accuracy term in the BIC or AIC is an acount for that? and where it is stored?






Gesendet von Yahoo! Mail.
Dem pfiffigeren Posteingang.