Dear all,
once again I need your opinion and valuable help.
I have built a logistic regression model which
predicts a certain dependent variable Y1.
Now, I want to calibrate my model to the prediction of
a dependent variable Y2 without having to build the
model from the start. Basically, I just want to
produce a mapping from my model Y1 to Y2. Y2 is
correlated to Y1 and it makes sense assuming such
mapping exists.
I am planning to do so by building a regression model
on the output of my original model (Y1) but using as
event Y2. This should basically calibrate my original
model to the prediction of Y2 i.e. Y2=f(Y1).
What do you think about my approach? Can you see any
problems I am missing out? How can I guarantee
robustness of my results, apart from validating the
model?
Thank you very much in advance for your help
Regards
Piero Bassu
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
|