Sorry if this is confusing! This visualization is a Dorling Cartogram, which sacrifices geographic location accuracy in order to not have overlapping points (the East Coast would be a tight unreadable bundle otherwise, as well as the Bay Area). The basemap is intentionally left as just an outline to provide some visual context.
I understand and accept that the location is not accurate to the map, but what was so confusing was that the locations didn't seem to be accurate even with respect to other locations, and that the location wasn't deterministic. That doesn't happen on the nytimes example.
If I toggle back and forth between sorting by population and sorting by Rent Price, the bubbles bounce around and end up in... some random arrangement. Sometimes it's reasonable, and sometimes it's bizarre.
Los Angeles is SouthEast of San Diego, Oakland is West of San Francisco and San Jose is Northwest of both of them. It's hard for me to avoid being pulled out by that.
That's definitely the map's weakness. We are using a force-based model which is nice in that it is dynamic (if lazy), but does not have the same consistency as the NYTimes example. For the NYTimes I understand they baked and tweaked the layouts in advance in order to play it as an animation as well as keep the layouts in the same position, which we should probably do on future vaguely geographic explorations like this :)
Sorry if this is confusing! This visualization is a Dorling Cartogram, which sacrifices geographic location accuracy in order to not have overlapping points (the East Coast would be a tight unreadable bundle otherwise, as well as the Bay Area). The basemap is intentionally left as just an outline to provide some visual context.
A great example of this in action is the Olympic Medal count visualization by the NYTimes: http://www.nytimes.com/interactive/2008/08/04/sports/olympic...