Logistic Regression Model Assessment

Riya

New Member
#1
I have run logistic regression on data. The concordant percentage is coming out 45. That is too low. However, Hosmer–Lemeshow test is coming out insignificant. It means we cannot reject the null hypothesis stating model fits the data well. That's good! The ctable also shows 70% of the data correctly predicted. My question - How contradictory results of concordant and hosmer-lemeshow test should be considered? How should i interpret this model?
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
Shouldn't the concordant percentage and the 70% be the same? You want the H-L to be insignificant to provide evidence toward good (model) fit.

I think there is a rule about certain percentages of observed and/or expected falling into the top deciles of the Hosmer-Lemeshow test that helps interpret it.