Nice! I really like building stuff with D3. I just put this up the other day though I'm still toying around with it. It's a visualization of what precincts are ticketing drivers in NYC. D3 has made doing stuff like this pretty easy.
Yeah, this map shows something slightly different (and possibly more interesting) than population density/grouping. Really it kind of shows which counties should be split up.
Compare the Houston and Dallas metro areas - Houston has a very distinct large bubble, while Dallas has more overlapping smaller bubbles which make it look smaller, even though it contains more people.
Usually you want to divide your visualization field by some other field. In this case, you might want to do (population / area). For other sorts of things like visualizing a particular demographic, you would often want to divide by population.
Perfect timing, just learning how to make a map in d3 for Bower stats [1]. Thought after experimenting with choropleth maps [2], bubbles on maps and dorling maps [3] [4], dorling maps appear to be the best way to show very dense info on relationships with vague references to geography.
I'd say the choice between choropleth and dorling maps depends on what sort of data you want to present, or am I missing something?
What I mean is, if the bubble size represents a different information from the color depth (i.e. figure 10, page 12 f the dorling paper shows population as a bubble, AFAICT) than it makes sense, but if it's just another encoding of already represented information it seems weird.
I.e. in the "dorling world" map I have lost plenty of informations (I have no idea what northern and eastern europe countries are what anymore, while I can tell them at a glance on a normal map) but I have not gained anything.
Just yesterday I watched Mike Bostock remake this example live at Edward Tufte's seminar in San Jose.
In the talk, he emphasized the importance of incorporating Makefiles into the workflow more than the text of this tutorial does. It basically documents exactly where you got the data from and what basic transformations you applied to it. The NYT uses this to streamline their process of developing graphics--so if, say, a graphic from six months ago needs to be revisited again for a new story on the same topic, nobody has to fumble around trying to remember where the data came from, etc. I got the impression that this was the advice of someone who had been burned by not following it in the past.
A lot of people are starting to use org-mode for that kind of thing. Fully reproducible research in a single file. To check the work, just export the file to pdf/html/etc., or "C-c C-c" the code blocks as you read through in org-mode. It's fantastic.
18 comments
[ 3.0 ms ] story [ 49.0 ms ] threadhttp://uturn.wolvesintheserverroom.com/
Super cool visualization, and great that you put the code on GitHub for everybody to see.
For anybody interested in the code behind:
https://github.com/akilism/moving_violation_scraper
https://github.com/akilism/nyc-moving-violations
The east coast has a much higher density of small counties, giving the NY area a look very different from that around LA.
Compare the Houston and Dallas metro areas - Houston has a very distinct large bubble, while Dallas has more overlapping smaller bubbles which make it look smaller, even though it contains more people.
[1] https://shan.io/bower/ [2] http://bl.ocks.org/mbostock/4060606 [3] http://www.dannydorling.org/wp-content/files/dannydorling_pu... [4] http://www.jasondavies.com/maps/dorling-world/
I'd say the choice between choropleth and dorling maps depends on what sort of data you want to present, or am I missing something?
What I mean is, if the bubble size represents a different information from the color depth (i.e. figure 10, page 12 f the dorling paper shows population as a bubble, AFAICT) than it makes sense, but if it's just another encoding of already represented information it seems weird.
I.e. in the "dorling world" map I have lost plenty of informations (I have no idea what northern and eastern europe countries are what anymore, while I can tell them at a glance on a normal map) but I have not gained anything.
In the talk, he emphasized the importance of incorporating Makefiles into the workflow more than the text of this tutorial does. It basically documents exactly where you got the data from and what basic transformations you applied to it. The NYT uses this to streamline their process of developing graphics--so if, say, a graphic from six months ago needs to be revisited again for a new story on the same topic, nobody has to fumble around trying to remember where the data came from, etc. I got the impression that this was the advice of someone who had been burned by not following it in the past.
For python specifically, this video is a must watch: https://www.youtube.com/watch?v=1-dUkyn_fZA