Transforming an "Uninformative" Distribution

This Demonstration shows the effect of transforming a uniformly distributed variable. Naively we may choose to use the uniform distribution to represent a state of no information, perhaps as an uninformative prior for Bayesian inference. However, if is uniformly distributed, even simple transformations of may not be—in this example the phenomenon is illustrated with powers of . To give a concrete example, suppose represents the radius of a circle. Then if is uniformly distributed (we might say we were completely "uninformed" about ) then we know something about the area of the circle (the distribution of squared is not uniform)! This suggests that the concept of an "uninformative" distribution is not as simple or clearly defined as it first appears.

SNAPSHOTS

  • [Snapshot]
  • [Snapshot]
  • [Snapshot]
    • Share:

Embed Interactive Demonstration New!

Just copy and paste this snippet of JavaScript code into your website or blog to put the live Demonstration on your site. More details »

Files require Wolfram CDF Player or Mathematica.