Multidimensional scaling of REM album covers: FlagSpace revisited

December 11, 2012
By

(This article was first published on Robert Grant's stats blog » R, and kindly contributed to R-bloggers)

By way of following up on an old in-joke, and doing something constructive because I couldn’t get to sleep, I thought I would revisit the Flag Space plot that I blogged about a while back and learn how to do it. The details are here at R-bloggers and the code is here at Github. It is surprisingly simple. I actually only had 15 images (REM album covers, don’t ask), so I just saved them from Amazon. They were JPEGs not PNGs, but the readJPEG() function works in exactly the same way as shown in Github. Implementation from there is very simple indeed.

REM-PCA1

Of course, it means nothing. Maybe Dimension 1 is dark-light and 2 is colour-monochrome? The stress is 0.17 which I think is a pretty good fit to 2 dimensions. There are probably more meaningful ways of getting distance / dissimilarity matrices for raster images, and I’ll consider them next. Then at some point I would like to get to grips with the tuneR package, which might lead to the same plot with analysis of the actual music. I really mustn’t drink coffee in the evenings.


To leave a comment for the author, please follow the link and comment on his blog: Robert Grant's stats blog » R.

R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...



If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.