2

In one dimension it can be shown that the highest posterior density (HPD) interval is the shortest; I found a proof in Subjective and Objective Bayesian Statistics (Section 8.4) by S. James Press where a suitable Lagrangian is built and used to find the solution.

Does anyone know of a proof for the higher dimensional case?

Xi'an
  • 90,397
  • 9
  • 157
  • 575
Valerio
  • 123
  • 5
  • 4
    The proof is the same: take a subset out of the HPD and you need more volume to achieve the same coverage. – Xi'an Nov 05 '17 at 16:41
  • 1
    @Xi'an thanks! in the 1-d case, is your suggestion equivalent to the proof above? – Valerio Nov 10 '17 at 10:25

1 Answers1

1

Answered in comments:

The proof is the same: take a subset out of the HPD and you need more volume to achieve the same coverage. – Xi'an

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467