Comments on: Step Chart Without Risers

By: derek

derek — Thu, 05 Jun 2008 15:25:30 +0000

demaws, sometimes the data becomes a story about how big the instantaneous rises and falls were, and when they occurred. When that happens, it may be better to have a graph that's all risers and no steps, like a sort of time-lined waterfall graph. (a step graph without risers is a sort of quantitative Gantt chart) More subtle intermediate cases might benefit from risers and steps that have different formatting, as recommended in Edward Tufte's 1983 classic The Visual Display of Quantitative Information. The last chapter I think.

By: Jon Peltier

Jon Peltier — Sun, 25 May 2008 22:22:39 +0000

demaws – “Maybe I’ll just have to plot it myself and see how it looks.”

That’s exactly right. So much of charting is experimental. Partly to investigate relationships in the data, and partly to see how best to illustrate these relationships.

By: demaws

demaws — Sun, 25 May 2008 04:02:43 +0000

Very Nice! The step graph clearly has affordances that lets it communicate way more information about this particular dataset. Getting rid of the risers was a thoughtful touch.

I’m still worried about how the step chart would look if the data was much more erratic than gradual (i.e. lots of up and downs). Would it convey the trend just as clearly as a line chart then? Maybe I’ll just have to plot it myself and see how it looks.

Also got around to looking around your website. Interesting stuff.

By: derek

derek — Sat, 24 May 2008 12:42:50 +0000

It wasn’t x error bars; it was a proper scatter series, from a data block very like yours above. My mistake, that caused me to abandon the line series, was to have the gap row with a gap value, and a gap date. The line series then insisted on including the risers, forcing me to source a second series from the same data block.

I went back and put dates in the gap lines, and the risers went away, making a scatter series unnecessary. Thanks for that.

I would still want a second series for individually-formatted risers, although doubtless a macro can be written to do that on a point-by-point basis to a single series.