Two-sample tests

From GusWiki

variation on the Voronoi approach:

• Dwyer, Squire (1993) - A multivariate two-sample test using the Voronoi diagram

(available online, but the pages are backwards, annoyingly)

The above paper says that it's better to use the dual of the Voronoi diagram as your bins (you draw lines perpendicular to the Voronoi lines, linking the sample points).

• Zech, Aslan (2003) - A Multivariate Two-Sample Test Based on the Concept of Minimum Energy

presents a neat idea, where the x's and o's are analogous to positive and negative -ly charged particles: if they come from the same distribution, you will tend to have a stable (low-energy) cloud.

• Baringhaus, Franz (2003) - On a new multivariate two-sample test

This one isn't available online, but the abstract explains the idea: if the two samples come from the same distribution, then merging them won't affect the average distance.

• Henze (1988) - A Multivariate Two-Sample Test Based on the Number of Nearest Neighbor Type Coincidences

I don't understand this one (yet).

You might also want to play around with different distance measures (e.g. weighting the dimensions differently).