topic Re: Predicting values using nonlinear regression in SAS Programming

Predicting values using nonlinear regression

kpberger — Wed, 30 Dec 2020 20:26:59 GMT

Hi all. Most of the observations in my large data set have non-missing values for both X and Y, but some are missing Y values. I am looking to predict values for Y using values from X. X may predict Y in a nonlinear fashion. While the end goal is to predict Y for those missing it, I would like to also predict Y for the whole population.

This is easy to do with a linear regression - just plug in the formula generated by regressing X on Y to those missing Y. However, how do I do this using some kind of nonlinear regression or splines? Thank you!

Re: Predicting values using nonlinear regression

PaigeMiller — Wed, 30 Dec 2020 22:49:48 GMT

Are you using PROC NLIN?

If so, the OUTPUT statement gives you predicted values for any row in the data set which has all non-missing X variables, whether or not Y is missing.

Re: Predicting values using nonlinear regression

kpberger — Tue, 05 Jan 2021 00:20:09 GMT

Hello. I actually ended up using proc glmselect to fit a model with splines per this blog post: https://blogs.sas.com/content/iml/2017/04/19/restricted-cubic-splines-sas.html

It outputs a data set that has the predicted values.