Skip to main content

Posts

Showing posts with the label spotify

Generate highly personalized music data using Exportify

Spotify generates gobs of data about music.  Most people have seen the end-of-the-year data Spotify generates for each user about their listening patterns . Most people don't know that Spotify also generates a lot of data about individual songs. Some of it is straightforward: tempo, genre, length. However, Spotify also has its own niche way of quantifying songs: Danceability. Accousticness. Here is a whole list of their variables and descriptions from researchers at CMU:  https://www.stat.cmu.edu/capstoneresearch/315files_s23/team23.html What does this mean for a stats teacher? You have access to highly personalizable data sets, rooted in music, with gobs and gobs of variables for each song...or artist...or album...or year of release...or genre (like, so many ways to divide up your data).  For instance,  I created a data set with Spotify data for 1989 and 1989 (Taylor's Version) to teach paired  t -tests . How do Taylor's re-recordings compare to the originals?...

Paired T-tests (Taylor's Version)

Ok, more Taylor Swift data for you. DID YOU KNOW that Spotify collects buckets and buckets of data about each and every song it provides (see:  https://www.spotify-song-stats.com/about ) So, I downloaded this information for 1989 and 1989 (Taylor's Version). So I could test for any differences between the recordings. Like, with data, not with my feelings and emotions. Specifically with a paired t -test. I get it. The sample sizes are very small. However, the data is still interesting. It makes sense that the tempo hasn't changed. Like, she did slow down or speed up anything. And that is super NS with an itty-bitty effect size. It is also interesting that acousticness has decreased. These are more heavily produced versions of the same songs (IMO), and while this change didn't achieve significance, it is a moderate effect size.  ANYWAY, you aren't really here for this information. You are here for data to share with your classes, yes? I'm here to help you teach your s...

Daniel's "Most timeless songs of all time"

This article, written by Matt Daniels  for The Pudding , allows you to play around with a whole bunch of Spotify user data in order to generate visualizations of song popularity over time. You can generate custom visualizations using the very interactive sections on this website. For instance, there is a special visualization that allows you to finally quantify the Biggie/Tupac Rivalry. So, data and pop culture are my two favorite things. I could play with these different interactive pieces all day long. But there are also some specific ways you could use this in class. 1) Generate unique descriptive data for different musicians and then ask you students to create visualizations using the software of your choosing. Below, I've queried Dixie Chicks play data. Students could enter their own favorite artist. Note: They data only runs through 2005. 2) Sampling errors: Here is a description of the methodology used for this data: Is this representative of all data...