How exactly are average importances for each attribute calculated in SMRT. I would assume that for each attribute, the difference of the utility of the level with the highest level and the utility of the level with the lowest utility would contribute to Average attribute importance, but this does not seem to be the case. Attributes with high utility difference between levels have similar importance when compared to attributes with low utility difference between levels.
Any information as to how SMRT calculates this would be helpful.