Eric Bridgeford

I meant for *all* of the dimensions -- it seems like one has no way to decide how well ZG2 did otherwise post-hoc; e.g., you can't actually produce a full...

Very good call. However, I think there are a lot of tricks to get "surrogates", at least for what I imagine to be the "most common" case that will arise...

Sounds great; though the print out (if I'm understanding right) should be a 3-column; sample id (1:n, in the order passed in), subject, then ranks. Agree within vs between-individual discriminabilities...

NB: hyppo 0.3.0 does not suffer from this issue, and works successfully.

randomForest (I use that package and based my code for writing/training classifiers around their interface) uses predict.* S3 interface. see [randomForest predict code](https://github.com/cran/randomForest/blob/master/R/predict.randomForest.R). In randomForest, you can see there's a...

the advantage is that let's say I train a bunch of models using a bunch of packages (randomForest, glm, R-RerF) I can use the same `predict(trained_model, X.new)` for all of...

definitely no requirement, but it will make peoples' lives easier the more accessible the algorithm is :) S3 methods are super trivial to implement; it would probably be sub 4...

PRoduce cmaps that can be imported with all color combiinations we will want

- Qualitative: `cm.get_cmap('Paired')` - Divergent: continuous `cm.get_cmap('RdBu')`, discrete `cm.get_cmap('RdBu', nbreaks)` - Sequential: continuous `cm.get_cmap('Purples')`, discrete `cm.get_cmap('Purples', nbreaks)`

Nice fix @anrddh. Solved issue for me! Commenting just so I get updates on this issue :)