Thank you Jessica, this helps.
So, when I add ".nlml" to the gaussian process machine I get the negative log marginal likelihood values for each fold and PRT. Just to be sure, when I simply compare those (mean values of folds) marginal likelihoods of the different PRTs I'm finally able to claim which questionnaire is better predictable from brain activity data (although I don't know yet how to this exactly).
thanks and sorry for the barrage :)
mike
|