Hi! I just want to know how to find a relatively good cut-off point for binary logistic regression model. I haven't learnt anything on ROC curve. Is it reasonable if I just change the cut-off point to the event rate?
This is what I currently got.
event rate/user rate (proportion of users in the sample)=69.6
When the cut-off point is 0.5 (default by SPSS), my classfication table is
When the cut-off point is user rate, i.e. 0.7, my classfication table is
I think the second one is better, but is it reasonable if I change the cut-off value to 0.7?
Advertise on Talk Stats