Behavioral Data Analysis with R and Python

Chapter 11. Introduction to Moderation

One of the most gratifying aspects of combining the causal and behavioral perspectives is that things that may seem entirely unrelated under one of them can turn out to be the exact same thing under the other. More simply put, when you have the right hammer, a lot of things are indeed nails.

So far, we’ve used causal diagrams to understand what drives behaviors on average: if temperature increases by one degree, keeping all the relevant other variables constant, by how much do sales of ice cream in C-Mart stands increase? But very often, we’re not just interested in that grand average, and we would like to break it down further:

• Does that number apply equally to the stands in Texas and Wisconsin? If not, this means our data shows an opportunity for segmentation.

• Does that number apply equally for chocolate and vanilla ice cream? If not, this means that there is an interaction between temperature and ice cream flavor.

• Does that number apply equally at low and high temperatures? If not, this means that there is a nonlinearity in the effect of temperatures on sales.

The hammer we’ll see in this chapter is called moderation analysis by social scientists, and it will allow us to address these three types of questions in the exact same way.

In the first section after reviewing the data and packages for this chapter, we’ll do a tour of moderation and see how it can apply to a variety of behavioral situations. Because the math remains the ...