I would like to test several (multivariate --between 5 and 20 dimensions) depth measures empirically. I'm looking for a couple of test-benches (read distributions). These should be:
- absolutely continuous,
- have convex contours (if you prefer, level sets)
- have non independent components.
- Easy to implement (i.e. implemented in R or source code available on-line).
- defined in arbitrary dimensions