Dot distribution maps of world population

[saga]

Is there anyone making decent dot maps of world population? It seems like it should be out there, but the maps I've found use a non–equal-area projection and/or let dots overlap each other, which are obvious dealbreakers for this type of map. Is there anyone doing this properly?

[Atarimaster]

but the maps I've found use a non–equal-area projection

I agree, that IS a nuisance!

and/or let dots overlap each other

Ummm. Are we thinking about different things here?
To me, a dot distribution map looks like this, and there, overlapping dots are not a dealbreaker but impossible to avoid in densely populated areas.

[saga]

Dots overlapping each other makes it impossible to count the number of dots in a region. When dots fully blanket an area, very different densities can look exactly the same. Though I can only see the abstract, this article seems to investigate this and conclude that dot numbers get underestimated by readers and that reducing the size of the dots can help.

Dot overlapping can be avoided by reducing the size of the dots or increasing the quantity each dot represents. In the one you linked, it looks like each dot represents an individual library, so the latter option isn't possible. For a population map, the quantity represented by each dot is arbitrary, so this can be done. I also think displacing dots to avoid overlap could be a good idea, though I haven't seen this done before.

[Atarimaster]

Dots overlapping each other makes it impossible to count the number of dots in a region. When dots fully blanket an area, very different densities can look exactly the same.

That’s true, that’s very inconvenient. But still, I believe that avoiding overlaps is, regrettably, impossible.
Look at this map:
https://upload.wikimedia.org/wikipedia/ ... spanic.png

The dots already are so small that making them even smaller would not be a good idea. Increasing the quantity each dot represents would, most likely, make many dots disappear completely, showing many parts of the U.S. empty.
But still, there are overlaps.

[Atarimaster]

One more thought …

I also think displacing dots to avoid overlap could be a good idea

I completely disagree here.
That would, so to speak, move a lot of people to a place where they actually do not live. Which is, in my opinion, even worse than using a non-equal area map.

Edit: OK, on second thought – not "even worse", but "just as bad as".

[saga]

The dots already are so small that making them even smaller would not be a good idea. Increasing the quantity each dot represents would, most likely, make many dots disappear completely, showing many parts of the U.S. empty.
But still, there are overlaps.

This map seems to be done on a per-county basis, with the dot count in each county representing the relevant population of that county. If we stuck to that method, I agree, and in fact I think this problem is already happening in the map. For example, northern Wisconsin is empty, despite there probably being enough Hispanic people in northern WI in total to justify a dot or two. I think the core problem here is the fixed size of the map divisions used. A better option could be to dynamically size the subdivisions so that each cell contains one dot's worth of population, then place one dot in each cell. This approach would be harder to implement, but would be robust against this type of problem, no matter how much each dot represents.

That would, so to speak, move a lot of people to a place where they actually do not live.

It would be a distortion of one piece of information, the positions of people, in service of preserving another piece of information, the population counts. Moving dots far from their correct locations would be silly and unrepresentative, but I think a slight distortion here could achieve a good balance, especially if we somehow indicate how the dots have been moved. Another option is to show high-density areas at a larger scale using insets. That would ruin the sense of population density, but it would perfectly show both dot counts and positions.

[Atarimaster]

A better option could be to dynamically size the subdivisions so that each cell contains one dot's worth of population

I bet that then again you run into the problem of overlapping dots in heavily populated areas.

Moving dots far from their correct locations would be silly and unrepresentative, but I think a slight distortion here could achieve a good balance, especially

I doubt that it’s possible to move the dots in this way in, for example, Chicago or NYC in the map. I bet you’d have to spray the dots over a large area to be able to identify single dots. But of course I’ve got no way of knowing because – well, as you say, I cannot count the dots.

[saga]

True, dynamically sizing the cells doesn't immediately help the overlap problem, because the dot counts in populated areas would be about the same. What it would do is handle sparse areas better, which would enable us to make each dot represent more, which would help dense areas. Ultimately there's no point in speculating without actually trying it out; I'll work on an implementation when I get the time.

[Milo]

This is why I favor heat maps over dot maps. Each pixel is colored with a brightness proportional to the density of the thing being measured in the area represented by the pixel, as though it's the average of a lot of smaller-than-a-pixel dots. Dot maps are just an approximation that's sometimes easier to produce, and might work well for things whose density is very low over most of the map (such as the large regions with not even a single library).

The linked library map also works well if your goal is not to ask "how common are libraries around these parts?" but just "how far is it from here to the nearest library?". Though clearly, if you're in one of the regions where libraries are really common, you're going to need a zoomed-in map to figure out the exact distance.

(Incidentally, I can find two world population heat maps on Wikipedia, one of which is plate carree and one of which is Winkel Tripel... neither of which are equal-area. And both maps are several years old. If you want up-to-date information that's also sensibly formatted, well, I have no idea where to even start looking.)

[saga]

Heat maps are definitely good for showing which parts of the world are dense. What they're less good at is showing total population counts for regions, or communicating things like "most people live in Asia". Dot distribution maps can communicate these things (if there's no overlap!) either by counting or estimating visual mass.