Multidimensional Scaling

This example starts off by showing a plot of european cities. Each of the cities are positioned so that the distances between each of them approximate the distances between them by road. The source of the data comes from the “eurodist” dataset, distributed with R. The positions of the cities are calculated using multidimensional scaling. The cmdscale function was used to perform this task.

A consequence of multidimensional scaling is that distances between points are approximated and no longer represent their true distances. As a result, the distances between cities can be distorted from their true distances. Significant examples of distances which are inaccurate are Athens — Rome, and Cologne — Geneva. This visualisation demonstrates this effect by showing the true distances between a reference city and all other cities.

This example is an R-based re-implementation of a similar visualisation created by Simon Barthelmé at http://bl.ocks.org/4482115.

The following video demonstrates the mds example found in the sjpMScThesis package, which can be downloaded at my Master’s Thesis page. To run this example yourself, open R and run the following command: library(sjpMScThesis) ; thesisExample("mds")

This video is best viewed in full-screen.

Download

This video example is available in both MP4 (x264) and WebM formats: