Hello all
I'm using a logistic regression to predict
probability of survival. For the
validation procedure I'm using respectively
(randomly) 30% and 70% for
validation and predicition sets .
By looking at the H-L goodness-of-fit test, the
p-value was 0.003 .
But when I run the model a second time, the H-L
test gives p=0.05.
Is there any problem with H-L test?
My question is Do we have to run the model
several times and choose the one
with the best goodness of fit? Or do we just
look at -2LL ?
Any help?
Thank you.
Omar Bouamra.
|