Data Visualization - BESegal

Good explanation why stacked line graphs a.k.a. area charts mislead people.

A good combination of the theory behind using visual analysis instead of tables with concrete examples.

Jeffrey A.Shaffer is another good resource for data visualization best practices.

Great source of CLEAN data from Philadelphia's public and government agencies. Runs the range from College test scores, to fire and police data for Philadelphia, to election data and so on.

Good examples of clean and to the point graphs. Done by the CBO - Congressional Budget Office. Topics: Spending on Unemployment, Funding for Child Nutrition, Cash Flow of Highway Trust Fund, & New Residential Mortgages

Yes, this is my own example of one reason not to use pie charts.

Tableau updates this list of free data sources available on the web. Good source of free data.

Analysis. Insight. Action. The New York Times does some of the finest data visualization work. Its R&D Labs created "Cascade" as a data visulization tool that lets it track which stories on it's website get the most traction on social media sites.

This gives it insight into which articles and authors are hot at any given moment. This turns to Marketing and Sales Action when it uses the information to sell display ads at higher prices while the pages are hot in social media.

Here's a great example of using Social Network Analysis to visulize the interconnections of facebook friends. Notice how those connections replicate the political boundaries of countries.

The founder of the Data Visulization Discipline

This deck is long and covers many basic points made by Stephen Few, Edward Tufte and others. Yet, for people who want a check list of what to do and what not to do, it will serve the purpose.

Here's the Cliff's Notes version of the 176 page deck. (Be sure to read Stephen Few and Tufte and the other heavy weights in the field.)

The Key Take aways:

1. Avoid 3d Charts at all costs. They distort the data. p.7

2. Use pie charts sparingly (no more than 2 or 3 slices), if at all.

3. Replace red/green color scales w/ other colors.  Help the color blind. Other experts suggest Orange-Blue.

4. Try fixed segments for color scales before using continuous color gradients.

5. Sort bar charts by value. It tells a story p. 39-41
6. Stacked area charts can distort data. p 53-59
7. Lable directly whenever possible, rather than using keys. p. 60-62
8. Tell people the point of the graph. p.71-83

Charts are good examples how to tell a story with data visulizations. These time-series graphs, show LeBron's position on court when he shoots and scores or misses.

Also be sure to see the heat-topographic graph towards the end.

Jonathan Drummey's blog Drawing with Numbers is filled w/ great tips for using Tableau

Pie Charts and the few times they work well. Another good post from Robert Kosara.

First, create a square waffle graph instead of a pie graph b/c many people use pie charts incompetently. Here's a link to how to make them in excel
http://www.juiceanalytics.com/writing/solving-the-pie

Second, yse a pie chart only when all 3 of the following conditions apply:

1. You can define the entire set in a way that makes sense to the viewer; and
2. The main purpose of the pie chart is to show the relationship of the parts to the whole. (Don't use if you want to show the relationship of the parts to each other); and
3. You limit your slices to 5 to 7 total (including an "other" for the remainder parts.)

More evidence that the New York Times does some of the finest data visulization work around. This web site collects some of the best of the NYT's work.

Use this to see examples of what Edward Tufte calls "information dense" data visulization. Also note the frequent use of context. E.g. charts changes in U.S. Gross Domestic Product include the context of events like Oil Embargo's or the U.S. Deployment of its Armed Forces, or Stock Market Crashes or Bull runs.

Alan Smithee provides some great examples of when to use which kind of graph.

According to OkCupid the best predictor of whether someone will consider sleeping with someone on a first date is if that person likes the taste of beer.

OkCupid has some great data scientists. They do a great job of asking the right question, preparing the right analysis, and then communicating it well.

Here's one great example...

Stephen Few, publishes perceptual edge. He along w/ Edward Tufte are leaders in the discipline of data visulization