riatelab/statsbreaks

Handle divergent series of values

TomBor opened this issue ยท 1 comments

TomBor commented

One last issue on my list before holidays ๐Ÿ˜Ž and more and open discussion about how to handle a divergent series of values.
Take for example positive and negative values in the same series, or values below or above a reference value like an index in base 100.

Sometimes we want to specify this special value and use it as a break. Then we need to split the original series in two (below/above special value) and run discretization on each sub-series.

Is that a case statsbreaks should handle?
Is there other way to treat those cases?

What you're proposing is to be able to apply classification methods to subsets. I think this is outside the scope of the lib, but I'm open to proposals.