rinuboney / clatern

Machine Learning in Clojure
Eclipse Public License 1.0
67 stars 12 forks source link

Add nonparametric ranking correlation metrics #31

Open rnowling opened 9 years ago

rnowling commented 9 years ago

I need to use the nonparametric ranking correlation coefficients and statistical tests Kendall tau-b and Spearman in my PhD work. If these are in-scope for Clatern, I'd like to add an evaluation module with kendall-tau and spearman-r functions.

mikera commented 9 years ago

This sort of thing possibly belongs in core.matrix.stats, see:

https://github.com/clojure-numerics/core.matrix.stats

rnowling commented 9 years ago

It makes sense to me to have evaluation metrics in Clatern itself to make life easier on users. Such evaluation metrics could also be used in cross-fold validation or other model selection techniques as part of a higher-level API.

I would be happy to add the metrics to core.matrix.stats and just use add wrapper functions in Clatern if that benefits both projects and their users.

mikera commented 9 years ago

I guess my criteria would be that if it is a generic statistical function it is better to push it upstream into core.matrix.stats so that it is useful for other purposes.

If it is more of a specialised evaluation function I agree it makes more sense in clatern.

You can always make an alias if you need it in a local namepace, something like:

(def foo clojure.core.matrix.stats/foo)