Open
Description
Here are a few cases where we want to cluster (ordinal or continuous) values:
- box plots with a continuous orthogonal dimension The box mark could use bin/rect instead of group/bar for quantitative data #1330
- facets on a continuous dimension Automatic binning when a facet dimension is quantitative? #14
- a subset of ordinal values with an "Others" group A transform to consolidate ordinal values outside the top n into an “other” category, perhaps in conjunction with the group transform. #144
- unknown (a place to represent "N/A", Color legends should display scale unknown & extremes when applicable #651, Communicate information about filtered data points #493)
If “group” could understand clusters (mapping dimension x to a cluster), and if “bar” could unpack them (mapping the cluster to to x1…x2 if appropriate), maybe all these issues might be resolved at once.
In a sense, a cluster exists both as a category (as part of a discrete set of categories), and as a subset of the original dimension (point location x, segment [x1, x2], or discrete subset of an original categorical domain).