THE UNIFORM PRIOR AND ZERO INFORMATION: A TECHNICAL NOTE
Andrew D. Ker
Abstract
Reasoning about maximal performance of hypothesis tests is difficult when there is supposed to be no information about some of the parameters. A common technique is to place a uniform prior on unknowns. In a recent steganography application [1], this gave a nonsensical answer, and we give an exposition and resolution of the paradox here. This note is a companion to [1].
Details
| Affiliation |
OUCL |
| Month |
June |
| Number |
RR−10−06 |
| Pages |
9 |
| Year |
2010 |