Answers and comparisons¶

A graph answers a question about the data. In one graph the ticks can carry the answer, and across graphs a shared scale carries the comparison.

A graph, like a paragraph, is there to answer a question.Jean-luc Doumont, Trees, Maps, and Theorems: Effective Communication for Rational Minds (Brussels: Principiae, 2009), on graphs, where the list of questions opens the discussion. Doumont lists the usual ones: how items compare, how values spread along a scale, how two variables relate, how one changes over time. One level up sits the comparison of comparisons, how those answers differ across subsets of the data. The marks carry the data. The furniture around them decides how easily the question gets answered, and the furniture is the part vanzelfsprekend owns.

In one graph, the ticks¶

A default axis ticks at round numbers whether or not the data reach them, which answers no question in particular. Ticks that sit where the reader is looking answer the graph's own. The resonance peak ticks at the band's edges and at the peak's measured frequency, so the reader finds the resonance on the axis instead of estimating it between 17 and 18. The grand tours tick each race at its slowest, median and fastest winner, so the axis says what a caption would have. Each is one locator, FeatureLocator for the peak and SummaryLocator for the winners, set after apply, and the frame reads them like any other ticks. The locators how-to has the calls.

The spine's end can say something too. Under nice the spine stops at the last round tick inside the data, and in the warming scenarios the worst one runs past it: SSP5-8.5 goes, literally, off the chart. A spine carried on to 5 would have covered the line and said nothing.

The names are the last of it. A legend answers "which line is this" by a round trip to the corner. A name at the line's end answers it where the line is.

Across graphs, the scale¶

Comparing comparisons takes panels, and panels compare only on a shared scale. small_multiples frames a grid on one x and one y by default, so a line that fills its panel is fast and a line that sits low is slow, across the whole figure. The grand tours read across three races only because of that. Framed each by its own record, every line would fill its panel, and the question "which race is fastest" would have no answer on the page. The scope, "figure", "column" or "row", is the choice of which comparison the reader is meant to make, and the small multiples how-to has the three.

The seam¶

None of this the library can choose. It cannot know whether the peak or the band matters, whether the slowest winner is a story, whether three races are to be compared or read alone. The locator, the anchor and the scope are yours. The library's part is that once you have chosen, the frame, the ticks and the labels make the answer legible without a caption. Frame and meaning draws that line in full.