There’s a challenge at FlowingData.com to remake the graph below. First, let me take a stab at explaining what is being shown here.
Start with the far right line. It shows the obesity rate for all people born between 1926 and 1935. We’ll call this group of people the first cohort. Notice the vertical axis shows per cent obesity, and the horizontal axis shows the age of the cohort. If we look at the first (leftmost) data point in the line, we see that about 13% of the cohort was obese when the cohort was 30-39 years old. Doing the arithmetic on the ages, this data point must be for 1965 (people born in 1935 were 30 years old, people born in 1926 were 39). The next data point shows per cent obesity when the cohort was 40-49 years old (1975), and so on for this line.
The next line to the left shows the same thing, for people born between 1936 and 1945, the line after that is for the cohort born between 1946 and 1955, and so on. Notice that each new line has one more data point on the left (until hitting the lowest age range) and one fewer on the right.
Each new line has roughly the same shape as the previous, and generally lies above and to the left of the previous. Lying above the previous line means that, in general, each cohort has a higher percentage of obesity than the previous. Lying to the left of the previous means that each cohort tends to get obese earlier.
It might be easier to understand the graph if the lines are revealed and explained one by one, than if presented all at once. In fact, the graph originally appeared in a presentation where the lines were presented one by one, and thus not as difficult to understand. (This was pointed this out in the comments by Naomi B. Robbins; I mentioned her book recently.)
Thus, I remade this graph as a sort of static version of a moving presentation. My semi-final version: pdf. I break the graph into eight subgraphs, one for each line, highlighted in black. The subgraph for Each subgraph shows the other lines as well, but muted, in light gray. This allows the trend of one cohort to stand out within the context of the others. If you move quickly along the subgraphs, you can get a kind of animation effect of the changes across cohorts.
I’m not yet satisfied with the axes. Between any 2 x 2 block of subgraphs, there is empty space that can be a little distracting once you notice it. (Tufte mentions this affect in one or a couple of his books.) I made a few variations trying to make this space less distracting (pdf), and I made a version with fewer tick mark labels. I would have to do some more experimenting to get a final version, and I’m not wanting to put any more time into this at the moment, so I’ll go with the semi-final version.