Functional Form

So far, we have mostly worked with very simple linear models. Here we introduce two common ways to make them a little more flexible.

First, we look at models which use logarithms. Depending on whether we take logs of the dependent variable, the independent variable, or both, the interpretation of a regression coefficient changes – from a change in units to a percentage change.

We then introduce interaction terms. These allow the effect of one variable to depend on the value of another, rather than being the same for everyone.

For now, the aim is simply to become comfortable with how these two ideas work and how their coefficients should be interpreted. We return to broader questions about choosing between different models later.