Friday, July 19, 2013

Visualization Friday: Probability Gradients

I'm fascinated with varieties of uncertainty -- ways of representing it, reasoning about it, and visualizing it.  I was very tickled when I came across this blog post by Alex Krusz on the Velir blog.  He presents a neat improvement over "box and whiskers" plot for representing uncertainty or variation in data points which he calls "probability gradients".


  1. The top graphic especially made me think of the work @eagereyes did:

    With the basic premise of "When visualizing uncertainty in data, a common choice is to use blur. While that may seem natural, it is unfortunately ineffective."

    1. Great comment and link, Jay! Robert Kosara's (@eagereyes) analysis (from his PhD dissertation) is really strong -- blur is not a good way to visually communicate a dimension of information related to uncertainty. He also says that blur is useful when you want to let other things (rendered sharply) stand out.

      When I saw the "probability gradient" experiments by Alex Krusz, My initial reaction is that it was a useful way to distinguish between distribution functions (i.e. Normal vs. uniform) rather than as a way to communicate the variance with any precision.

      But, I'm also interested in using blur in other settings to communicate a specific variety of uncertainty -- vagueness -- but this is different from Alex's examples which are focused on variability.

    2. One more thing -- In the examples above, Alex uses color lightness to create the blur effect, so he's not depending on blur alone. The data with narrow error distributions appear darker. The "blur" part is mostly informative about the tail, compared to the sharp rendering of the uniform distribution.