2

Are there distributions where no conjugate prior exists?
Is there a necessary and/or sufficient condition which guarantees the existence of a conjugate prior?

Edit:
Why has this question been closed? None of the two questions which are supposed to answer this question actually answers this.

helperFunction
  • 175
  • 6
  • I think that conjugate priors exist only for distribution in the exponential family. (I am sure that any exponential family distribution has a conjugate priore, but I'm not as sure about the other way..) – Pohoua Nov 27 '20 at 10:13
  • See https://stats.stackexchange.com/questions/192554/aside-from-the-exponential-family-where-else-can-conjugate-priors-come-from/192675#192675 – Christoph Hanck Nov 27 '20 at 10:25
  • @ChristophHanck I believe that question answers the following: "membership of an exponential family is not a necessary condition for the existence of a conjugate prior". Is there a sufficient condition? Or a distribution that doesn't have a conjugate prior? – helperFunction Nov 27 '20 at 10:30
  • 1
    I am not sure what the consensus among really knowledgeable Bayesians is here, but such a result may be complicated by the fact that a lot hinges on what how the "family" is defined to which priors and posteriors must both belong for the prior to be conjugate for a given likelihood. If you define it to be the family of all probability distributions, then a conjugate prior exists, but it is of course not a very interesting result. – Christoph Hanck Nov 27 '20 at 11:32

0 Answers0