How set a RuleX and RuleY at mean/median of the data?

[nimitnshah]

Hi!

I've been trying to get a RuleX position on the axis at the median value of the dataset. If we take the penguins dataset for example, and I put culmen_length_mm on the x axis, ideally I'd want a single RuleX line denoting the mean of all culmen_length data values. I tried using the Group transform but it outputs multiple lines across the axis, maybe because Group was meant to be for ordinal data? If it's easier, a RuleX on midpoint of the axis would also do. I feel like this should be easy to do but I'm missing something.

[mbostock]

Not sure what you tried (posting code is helpful for clarity!), but here’s a fresh example for you:

Plot.plot({
  marginTop: 18,
  marks: [
    Plot.tickX(penguins, {
      x: "culmen_length_mm"
    }),
    Plot.tickX(
      penguins,
      Plot.groupZ(
        { x: "median" },
        {
          x: "culmen_length_mm",
          stroke: "red",
          strokeWidth: 2,
          inset: -8
        }
      )
    ),
    Plot.text(
      penguins,
      Plot.groupZ(
        { x: "median", text: "median" },
        {
          x: "culmen_length_mm",
          text: "culmen_length_mm",
          frameAnchor: "top",
          dy: -18,
          fill: "red"
        }
      )
    )
  ]
})

https://observablehq.com/@observablehq/plot-median-rule

[nimitnshah]

Oh sorry, I thought I had posted the notebook! Here it is: https://observablehq.com/d/a613accffa2bd853. In the notebook, ideally I want a RuleX which would draw itself at the median of X data, and a RuleY which would draw at median of Y data. Also, is there a reducer we can use to draw it at the midpoint of the axes' (is there a way to access d3.extent/2) ?

Why are we grouping on Z instead of X? Does Z take overall data as sample, whereas GroupX would take each data point as a single datum? Sorry if it's a stupid question, just trying to understand the ins and outs of the Group Transform I guess?

also, THANK YOU so much for everything you and your team do in creating and supporting all of this! You've changed dataviz for life for me :)

[mbostock]

Thanks for posting the notebook, and glad our tools are useful for you. 🙏

“Grouping by x” means “generate a distinct group for each distinct value of x”, which wouldn’t be helpful here because nearly every x value is distinct: you’d get as many rules as you have input data. So we’re grouping by z (otherwise known as color) instead — and when there isn’t a z dimension, everything gets put into a single group.

Plot.plot({
  inset: 10,
  color: { legend: true },
  marks: [
    Plot.frame(),
    Plot.dot(penguins, {
      x: "culmen_length_mm",
      y: "culmen_depth_mm",
      stroke: "species",
      tip: true
    }),
    Plot.ruleX(penguins, Plot.groupZ({ x: "mean" }, { x: "culmen_length_mm" })),
    Plot.ruleY(penguins, Plot.groupZ({ y: "mean" }, { y: "culmen_depth_mm" }))
  ]
})

As for a reducer that returns the midpoint of the scale domains, no, there isn’t a built-in reducer that does that. But you can write a custom reducer that does that, e.g.:

Plot.ruleX(
  penguins,
  Plot.groupZ(
    {
      x: (X) => {
        const [x1, x2] = d3.extent(X);
        return (x1 + x2) / 2;
      }
    },
    { x: "culmen_length_mm" }
  )
)