Model assessment in cummulative logit Regression

[lionking19063]

Hi,

I am building cumulative logit Regression model with "PROC LOGISTIC". I wonder how to evaluate the performance of the  model. I saw c statistics in the output, may I use that one as we use it in the binary logistic regression (the higher the better)? Thank you.

[StatDave]

Yes, you can use that statistic (the concordance index, c) for comparing competing models. See also the information in this note on assessing fit in multinomial models.