Quantile-based discretisation function

[JosephBond]

Fluid library function to bin a set of “observations” into a set of (potentially non-uniform) bins.

As a test case, use the new function to compute the 5-95% interpercentile range from the sample we generate from the preprocessing step in explorable-viz/transparent-text#22. This will make up part of the pipeline for the examples, and is needed to provide a data source for the whisker plots on the bar chart bars, and also for computing the appropriate probabilities to use in text like “very likely”.

Python libraries to consider:

• pandas.cut (supports custom bins)
• pandas.qcut (quantile-based bins)
• numpy.digitize (similar but doesn’t require bin labels)

Going forward let’s use Python-inspired names for library functions, to leverage #1139.

See also:

• explorable-viz/transparent-text#26

[rolyp]

@JosephBond Added some clarification to this task and renamed from “Calculate Interpercentile Range from empirical distribution”.

[rolyp]

@JosephBond It looks like qcut does take an argument that allows you specify the target quantiles, so we could take a similar approach. E.g. something pandas.qcut(xs, q=[0, 0.05, 0.95, 1.0]) but without the named argument syntax. Renamed task again.