14 comments

[ 4.9 ms ] story [ 40.0 ms ] thread
In case you were wondering: "About the Visualizations The visualizations in Data USA are powered by D3plus, an open-source visualization engine that was created by members of the Datawheel team."
Hi everyone, I'm Dave, one of the developers of D3plus (and Data USA itself). Thanks everyone for checking out our site. If anyone has any questions feel free to shoot them my way.

http://d3plus.org/

It is very snappy on my browser, especially for the amount of information that is being presented. How are you using d3plus, build tools, and local storage solutions (or service workers?) to accomplish this?

I haven't explored it too much yet, but I'm enjoying the breadth and succinct presentation of all this information in one place.

We use a very aggressive page caching, so that all of the text and statistics are pre-rendered when the page is requested.

The visualizations only initiate when they are visible in the window. When that happens, they make an API call for the data and then initialize the D3plus visualization.

All of the visualizations use shared attributes, for things like location names and industry colors/icons. Those API calls are stored in the browser using localforage. https://github.com/mozilla/localForage

Hi Dave, I was looking at this page http://datausa.io/profile/geo/pennsylvania/#housing I noticed that a lot of the ranges on the x-axis are not consistent. Is there a reason for that?
Jon here - also worked on Data USA.

Several of the visualizations use data from ACS summary tables and the axis in several of the visualizations reflect the underlying buckets provided by the Census Bureau.

Really annoyed by the tooltips for some reason in most JS visualizations. It is fairly easy to tell what the exact numbers are already.

I think it would be a nice touch to make the standard error bars pop out or something else interactive. On printed plots, error bars is perhaps the only thing that you both want and don't want that would benefit from the dynamic nature of these.

Agree, the visualizations should definitely have margin of error incorporated into the designs.
One of the stats they show is gini coefficient. Is There anything counterintuitive about that statistic that I should keep in mind? For example, do high income areas have a high gini (simply because of the construction of the statistic)? or maybe small areas tend to have a smaller gini?
GINI is a measure of the spread of a distribution so even if in theory everyone earned $1,000,000 it could be seen as "equitable" according to GINI. What matters in GINI is what portion of total amount are distributed to what "buckets". So high/low income doesn't necessarily impact GINI per se. Checkout https://en.wikipedia.org/wiki/Gini_coefficient
Is there a knowledge graph being built in the background - powering the project ?
Beautiful! I like looking at the ratios for cities, patients to Primary Care, Dentists, Mental Health etc. SF doesn't score to well :-(