A Version of Quadratic Regression with Interpretable Parameters

The quadratic regression model is popular and effective in describing a wide variety of data, but it is based on a function whose parameters are not easy to interpret. We suggest an alternative form of the quadratic model that has the same expectation function, but also has the useful feature that its parameters are interpretable. Examples are provided of a simple regression problem and also of a nonlinear mixed-effects model. The models can be estimated with available software.