Finding the right curve can be tricky. The image is CC by Martin Gommel.

In my previous post I wrote about the importance of age and why it is a good idea to try avoiding modeling it as a linear variable. In this post I will go through multiple options for (1) modeling non-linear effects in a linear regression setting, (2) benchmark the methods on a real dataset, and (3) look at how the non-linearities actually look. The post is based on the supplement in my article on age and health-related quality of life (HRQoL). Continue reading →