ROC Curve for decision tree with categorial dependent variable?

[reventon]

Greeting,
      I'm proposing to compare numerous prediction models using AUC scores. The models are three decision trees based on the CHAID, Exhastive CHAID, and CART algorithms. I have two dependent variables the first (categorial with five categories) is a categorial variable and the second variable is dicotomous. I understand that ROC curves are 2x2 tables, is there any way to compute a ROC curve for the categorical dependent variable? Or do I have to limit the analysis to the dicotomous dependent?

Thanks,
    Reventon