A good visualization + some comments about good visualization: Obama’s 2011 Budget Proposal, Department by Department

Another pat on the back for the NYTimes graphics department (or whatever they call it internally - I'll let my friend @GrahamYRoberts correct me) for once again showing us that a well-designed visualization is better than a million words or a thousand statistics.

The other day, I read a statistic that "70% of employers have rejected job applicants over internet behavior." I don't mind the information, Guy, but there are at least 5 ways to mis-interpret that, given how bad people are at statistics, reading, and logic:

  • "70% of employers have fired someone for internet behavior" 
  • "70% of job applicants are rejected for internet behavior" 
  • "70% of people have a trashy/slutty picture on Facebook" 
  • "You only have a 30% chance of getting a job if you are on the internet" 
  • "I have a 100% chance of being unemployed."

(To be fair, Guy highlights the fact that 86% of HR folks said that a positive digital identity can have a favorable impact on an applicant's chances at the end of the post.)

New rules:

  1. Statistics can't be used as headlines / tweets without supporting info. 
  2. Statistics can't be used AT ALL if they can be easily misinterpreted in 3 or more ways by a high school educated person (a level that is lower than you think).

Further, statisticians need to do a better job in visualizing their data. There are some great researchers doing this out there, but the majority of us stink at it (I include myself, but I'm trying to get better). There are lots of tools out there (I'm told R is brilliant for this), and the success of a paper can pivot on it.

I just read a really good paper that was an absolute bear to read because of how the statistics were presented. One of the tables presented a set of ratios in bold with a set of inverse ratios (for different conditions) in regular font. How does that make any sense?

We're getting closer to understanding why people are bad at these things (beyond the intuitive observation that they are hard). In the meantime, quick and dirty statistics can drown along with newspaper companies that run them.