Comparisons of possibility-and probability-based classification: an example of depression severity clustering