A little background,

The dataset I've been given (to hopefully analyze) where a group of people have coded 288 music videos for the presence or absence of 88 different variables (e.g. band members wearing flannel shirts, or song is in verse/chorus form). Each row is a song, and each of the 88 columns has a 0 or 1, indicating presence or absence of the variable, respectively. The dataset looks something like this:

SongName . Var1 . Var2. ... Var88

SongA . 0 . 1 . ... 1

SongB . 1 . 0 . ... 0

SongC . 0 . 0 . ... 0

I was told to "explore relationships" between these variables, but I am unsure of how to proceed. There is no dependent variable (all of the videos are from a similar group of songs).

I originally thought about some sort of massive chi-square test, or calculating chi-square for each combination of variables, but neither of these seem right either.

Any help on ideas of how to proceed would be very helpful! Thanks!

  • $\begingroup$ I guess I would first try to detect if any of the 88 variables have identical patterns and than continue with either pca of efa to identify some categories to get a basic understanding of the data $\endgroup$ – Mr Pi Oct 16 at 23:26

You could think of this kind of like a community analysis. You have a bunch of different "sites" (i.e. bands) which each have a unique collection of "species" (i.e. clothing, song type, etc.). When you look at it like this then it becomes quite easy to try and pick out trends in the data by using dimensional compression techniques.

There are some quite good and easy to use packages in R to produce these quickly and easily and give you some easily interpreted plots - see the vegan package and it's metaMDS and vegdist functions. There are also quite a few helpful tutorials kicking around the internet to carry out these kinds of analyses.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.