Plot.movingAverage could generalize to other rolling aggregations.

[mbostock]

E.g., https://www.rdocumentation.org/packages/zoo/versions/1.8-8/topics/rollmean (mean, max, min, median, sum).

[Fil]

and array-blur (which has the advantage of keeping the full domain)

[Fil]

Geometric moving average, exponential moving average… are also in use. Since we don't want to cover every statistical corner, it'd be nice to have an entry point which would feed the (z and facet selected) series into a user-defined algorithm.

[mbostock]

I was thinking we should add d3.rollmean, d3.rollmin, d3.rollmax etc. to d3-array, and then consume them here?

[Fil]

Yes I think it's the correct move. I realize that I'm often using the exponential moving average (though I just discovered yesterday it was called like this). Keeping just one value + a half-life multiplier is very powerful.

[mbostock]

Yes, that would be great too! 👍

[Fil]

moving the issue to d3/d3-array#192 ; for Plot we just need to write a generic entry point (and keep the basic rollmean baked in, probably?)

[mbostock]

Added in #245.