[JBT]

One more vote for two- and four-axis graphs.

Thank you for putting a paper together on this topic. I agree with the attractiveness of small multiples and of the frequent confusion of two- and four-axes graphs, but I also share other commenters' views that your paper is prescriptive.

The Economist magazine uses two vertical axes graphs with great precision, clarity, and compactness. Their trick is to use a common grid line and color the axes labels with the color of the data represented on them (see, e.g., 8/28/10, p. 61). Oh, that OmniGraphSketcher could do that! The eye spends no time learning the lay of the land.

Could a small multiple do as well in the same space? I'm not so sure. You'd introduce another horizontal axis (reducing your data-ink ratio), or introduce a single-axis interpretation issue (say if you "stacked" the data in two regions on the single y-axis), or introduce a competing scales issue (e.g., even stock prices in dollars and volumes in millions may have vastly divergent dynamic ranges).

Are correlation plots a perfect substitute? I agree with the paper that sometimes they are the preferred option because correlation is the analytical issue at stake. However, they mask time, and that may be a high-stakes variable. Exploration may require being able to observe, "These series diverge in the late 1990s and converge in the early 2000's."

Will users aesthetically abuse the second-axis capability if it exists? No doubt. But that's their responsibility not yours, especially if you built in a default to color the axis labels by the data series color and plot on a common grid.

By the way, kudos to your development team! OGS is a DREAM to work with. It's got just the right mix of tweak-ability and speed and it's got very creative features I've wanted for decades (like shading under a curve). OGS reinforces my decision to switch to Mac every time I use it. It Just Works.

Now, when will you jump into the GIS space?