Help with non-linearity (log odds and predictor) in logistic regression - next steps

#1
Hi,
I posted this previously in the wrong section of the forum . . . I think it belongs here.
I'm new to modelling and logistic regression, so I need a bit of help. I'm building a logistic regression model, and one of my continuous variables violates the linearity assumption (log odds). I can't categorise the continuous variable because I don't have enough events, so my only options are to add a quadratic or higher polynomial term, or fit splines. However, I have no experience with either, and I'm working with SAS and would have no idea what to enter into the code. As I mentioned before, I'm still learning, so reading online has been a bit of a challenge because the language is a bit advanced for me. Any help would be greatly appreciated. I've attached a picture of my data, from when I was first checking to see if it was linear or not.
 

rogojel

TS Contributor
#2
Re: Help with non-linearity (log odds and predictor) in logistic regression - next st

hi,
in principle there is nothing special about adding nonlinear terms - but you will need to be a bit more careful about selecting the right model, i.e. chosing a good selection criteria. I would start with adding quadratic terms and only move to more involved models if it is really necessary.

regards