Logistic Regression Research: 97% Concordance, No Sig Variables

1 reply · 1K views · last reply
mathmathRW
Messages
8
Reaction score
0
I am doing an independent research project and I have written a logistic regression program in SAS. The percent concordance is 97%, but hardly any variables are significant. Can anyone help me understand why this would happen?
 
There are many possible reasons for this. Would you be able to describe the situation in more detail?
If all of the significance is in only a few variables, then you will not have many significant variables.
Otherwise, you may have some dependence issues, and the program can't determine which variables are better than others.
I would try reducing the model, and then applying some stepwise regression techniques to find the most significant factors.