07-12-2011 06:15 PM
I have build a logistic regression model (binary response) and having trouble using the model to calculate the estimates of the linear predictor (log odds) for each observation.
I have some quadratic and cubic terms included in the model and my estimated intercept is about -700. After I plugged in all my numbers in the the model, I found out that my calculated log odds was
quite different from the log odds generated from the ''XBETA=" statement. Normally when I have no quadratic or cubic terms in my model, my calculated log odds are really close to the ones generated
So I'm just wondering how exactly the log odds is calculated in SAS in this case when i have terms wtih higher powers.