Explain how to change the scale for bivariate reference example

[MarcSkovMadsen]

I would claim the scale of the example in the bivariate reference example is not useful for "production".

By the way the scale is not explained. Me guess is that its some relative scale that sums to 1. But I am not sure as it is not explained.

If I was to hand this plot over to any one else I would have to explain the scale and probable change it to something easier to read for many people. Business managers would like something closer to small integers than tiny floats I would claim.

My request is to get the scale explained in the reference documentation and improve the example to make it useful by showing how to adjust the scale to something more readable.

[maximlt]

I think that indeed the bivariate plot would need to be documented more. Unless there is something wrong with how it is computed, I'm not sure there should be a way to change the scale, these are the densities computed by the underlying algorithm. What do you think @jlstevens ?

[jlstevens]

The Bivariateelement is pretty much entirely defined by this operation which computes a Polygons element.

I'm always happy to see the docs improved but as to whether the scale is defined sensibly or not, the main thing I would say is that the HoloViews statistical elements were mainly intended to match seaborn instead of deciding our own conventions. If the behavior matches how seaborn computes it, then I would say that is intended behaviour, and if it doesn't I would file that as a bug on the HoloViews issue tracker.

[philippjfr]

I would say that colorbars should not be enabled for bivariate plots by default because densities are basically uninterpretable.