Comments on non-contiguous-cartogram

Feedback
1

@mbostock I’m impressed by the number of high-quality notebooks you publish on top of everything else you contribute to the community. Thank you.

A couple of comments/ideas to consider that came up while looking at the following:
Non-contiguous cartogram / D3 | Observable. Nothing requires a response.

Meta:

  • I would love to be able to comment directly on a notebook. I find myself not commenting because there is no easy way to do so. I think comments can help us engage with each other’s work even more.
  • Maybe even commenting on particular cells and even particular lines or sections of a cell would also be of value

Comments about the post:

  1. I love the meaningful transition between 2008 and 2018 – people are getting more obese as shown by a growing state size – cool. However, I can’t get over the cognitive dissonance with the fact we use the geospatial area to represent population. States that are large in size but have low population density are overly represented.
  2. I found myself wanting to see which states changed most or least between 2008 and 2018 – any outliers/clusters?
  3. Obviously harder – but it would be perhaps more visually meaningful and visually appealing if the shrinkage mechanism based on contours rather than linear x/y reduction. Hawaii will be the hardest, but maybe with a minimal required contour diameter cutoff would help hide the really small parts.
  4. IMO, too much geo-detail distracts from the message
1 Like
2

We’re working on better public comments! Stay tuned. We hope to ship something in the next couple of weeks. (In the meantime you can send private comments via suggestions, but it requires forking the notebook and does not allow for public discussion.)

As for this map:

Area does not represent absolute population here; instead it shows the proportion of the state’s population that self-reported as obese. So for example in California in 2018, the filled area covers 25.8% of California’s area, and represents 25.8% of California’s population. Hence this cartogram allows meaningful comparison of the prevalence of obesity across states (although it would certainly be nice to have more regular and equal-sized samples for more detail).

That said, I’m not wild about non-contiguous cartograms, and I’m fairly sure this representation is worse than a simple choropleth: if you remove the color encoding from the non-contiguous cartogram, it’s practically impossible to compare states, which suggests that the area encoding of proportion is not very effective.

4 Likes
3

You’re absolutely right. I suppose it is the comparison that the reader naturally tends to when they are presented with all the states side-by-side – we know the data points are proportions but we forget that the size of a state is completely arbitrary because it is so visually dominant. I see this akin to the known problems with pie-charts – the human brain doesn’t compute comparisons of the parts very well – plus, in this case, the size doesn’t represent population which is important.