What makes for what we call graphical excellence? Well-designed presentation of interesting data, which is this combination of 3 things. Substance, like the historical knowledge, statistics and design. So, good figures have all of these things. Give the viewer the greatest number of ideas in the shortest time, with the least ink and the smallest space. So in some ways this is, you know, a little over the top. But it gives us things to push for. If you can say something in less space, use less space. If you can say something with less ink, meaning less print, do it! If you can say something, if you can put more ideas in the same space, do it! Make your presentation data-rich.
So, that’s the kind of idea. How can we enrich a particular presentation of data? How can we do it in less space? Most analysis, you know, are multi-variant. In other words, they’re not just a simple time series. And then again, there’s this transparency theme that keeps coming back and back. You know, again and again. You have to tell the truth about the data. You have to be consistent in how you present things and present data. You don’t shift units and do other things to tell a story. So there’s a kind of transparency and honesty that goes with great, great figures.
Because, once you abandon truth telling, and once you’re trying to trick the reader and lie with statistics, the whole point of the graphic is basically destroyed. Graphical integrity is this idea, this theme, that you really can’t create effective figures if you’re not being honest and truthful. One thing he points out, I didn’t put a figure of it here, is a lot of figures are made to look misleading by stretching the y axis. To make some variable look like it’s skyrocketing. So there’s a lot of, you know, politically motivated or ideologically motivated graphics that do that. That have that kind of weird shape.
You know, most graphs have this kind of relatively either balanced square-like shape or they’re kind of longer to the horizontal. Which he claims maybe is just kind of natural like we see the horizon. There may just be a naturally satisfying way of viewing data that’s slightly more horizontal than vertical. But a lot of misleading analysis have this really elongated kind of look. Another thing that folks do, instead of stretching the y axis, is they use nominal values instead of real values at times. Again, to create this kind of skyrocketing feeling, which could be used for political purposes. Like, “Oh the cost of!” you know? Or, “Oh the price of the price of something is going up so much.”
Well, maybe not in real terms it’s not going up that much. Principles of graphical integrity. Show the data again. But in particular, show data variation not design variation. This is actually a pretty interesting point. People, sometimes, will present their data, different pieces of their data and they’ll modify design elements in ways that could be misleading. You want to make sure that the variation across panels in a figure or across figures if you’re really being honest about your data, is really being driven by variation in the data. Not by these design choices. So, that’s a fundamental point. You of course need clear and detailed labelling. You don’t want there to be ambiguity.
There’s this principle of having figures and graphics be, to some extent standalone items. Items that you could sort of approach and get enough information about what’s going on in the figure. I’m not saying you throw away your article or you throw away all your other research. But, they have to be approachable on their own and you may need additional information. But you don’t want it to be just sort of dangling out there. So if someone looks at it, they have no idea what’s going on without rereading the paper for 20 minutes. So you want there to be detailed labelling and less ambiguity.
Representation of numbers in a graphic should be proportional to the quantity. So this is another thing that he gives a lot of examples on. Again we talked about standardizing measures, say of money or other outcomes in real terms or per-capita terms. And don’t quote data out of context. This is something that Tufte discusses, his point is, “You know, you don’t always need a graphic.” He’s not saying that you have to turn everything into pictures. And, in fact, his point is for some data sets, especially relatively small numbers, you absorb them just as well or better in a small table. So yeah, when there are complicated relationships; geographic data, time series data and multiple variables, etc. it has to be graphical.
I think the general principle is, for limited amounts of data, tables are good. For complicated and large amounts of data, figures may be better. That’s kind of the general point. Whom are we making data graphics for? How sophisticated should they be? Tufte’s point here, and I think the world has become more sophisticated, so this point is probably more and more relevant. He says, you know, we shouldn’t dumb down figures. We researchers shouldn’t do it, newspapers shouldn’t do it. Basically, people’s capacity to process complex data is much greater than most people think, or most researchers think or most graphic designers think. You know there’s this view, and again in economics, we have some of this.
The sophisticated stuff is in tables, or in our equations. And the kind of simple stuff is in the figures. Or the simple stuff is in a graphic for unsophisticated readers. You know, let’s have a simple summary figure. Well, if it’s that simple you don’t really even need a summary figure. The whole point of the figure is to bring in some complexity and some interesting pattern.