ABSTRACT
In Chapter 16 we'll build our first hierarchical models upon the foundations established in Chapter 15. We'll start simply, by assuming that we have some response variable Y, but no predictors X. Consider the following data story. The Spotify music streaming platform provides listeners with access to more than 50 million songs. Of course, some songs are more popular than others. Let Y be a song's Spotify popularity rating on a 0-100 scale. In general, the more recent plays that a song has on the platform, the higher its popularity rating. Thus, the popularity rating doesn't necessarily measure a song's overall quality, long-term popularity, or popularity beyond the Spotify audience. And though it will be tough to resist asking which song features X can help us predict ratings, we'll focus on understanding Y alone. Specifically, we'd like to better understand the following:
What's the typical popularity of a Spotify song?
To what extent does popularity vary from artist to artist?
For any single artist, how much might popularity vary from song to song?
