And also remember that this map is automatically produced based on your query, so having slightly irregular ranges, and a range for a single value could be forgiven for the fact that there is no human involved in the production of each map, and while it may work really well for one query (John Smith maybe?) it may not work so well for another. I guess this really is the difference between rapidly produced map for quick information, and a cartographic work-of-art :) or am I being too kind?

You do make some very good points. Maybe I won’t criticize maps when I’m sick, anymore, as it seems to impair my thought processes and encourage me to say dissent-generating things.

I actually usually disagree with the whole normalize-by-population thing, as you and Zach do. But in this case I was willing to repeat the party line – though I did toy with the notion of making an argument that they should show the population density of David Wilkinses. Maybe it’s because I just found it boring, which is an unfair criticism. What I think might be better, actually, would be to put a normalized and un-normalized map side-by-side on the page.

The classification scheme I don’t have a problem with, apart from perhaps the poor labeling of the single-member highest class. It’s just an equal interval classification. The class ranges aren’t even because the whole range of 1-53 isn’t divisible by 5. Hence three classes of 11 and two of 10 (11-10-11-10-11). Granted, equal intervals don’t provide a clear picture of the subtleties of the spatial distribution of a phenomenon, but they do give a pretty unbiased look at it. There is some value in seeing that the value is very low in a lot of the places and pretty high in just a couple of spots. The trickiness of finding a balance is well illustrated by trying to map world countries by population: you can either have 80% of countries in the lowest class, or you can distribute them more evenly but entirely lose the fact that India has, like, 800 million more people than the next most populous country. (That’s oversimplifying the options, I acknowledge.) How you classify it depends on what you’re trying to convey.

Normalizing the data is another point I would dispute, though to do so is to be a bit more of a carto-rebel. Normalizing data, as you know, is a pretty strong rule in choropleth mapping, but I have come to agree with our pal Zach Johnson’s challenge of that notion, namely that normalizing by area is the only thing that makes sense as a hard rule. If not area, then it’s fair to toss the normalization rule out the window if we want. Without normalization the map here does deliver on its promise to show the “Distribution of David Wilkins Listings across the United States,” and in fact to normalize it by anything other than area would actually be showing something else. But this is opening a whole different can of worms. Talk about thinking too much!

All that said, I’d bet large sums of money that the developers of these maps didn’t give a second of thought to those issues. Maybe it worked out for them… or maybe it didn’t.

I’m actually going to add a bit right now about the classification scheme, since you’ve reminded me of another few things to talk about…

Or am I thinking about this too much?