And if so, could somebody give me a concrete example?
$\begingroup$
$\endgroup$
2
-
3$\begingroup$ It depends on what you regard as noninformative, and in some cases whether you require the prior to be proper. To take the example of a binomial likelihood, the conjugate distribution is a beta distribution, and some people regard a uniform Beta(1,1) as uninformative, some might use a Jeffreys Beta(0.5,0.5) prior, and some a Haldane improper Beta(0,0) prior. But since these are different, others might argue that simply choosing one would be informative $\endgroup$– HenryJan 11, 2022 at 9:14
-
1$\begingroup$ @Henry you should make it an answer. $\endgroup$– Tim ♦Jan 11, 2022 at 10:20
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
As requested in comments:
It depends on what you regard as noninformative, and in some cases whether you require the prior to be proper.
To take the example of a binomial likelihood, the conjugate distribution is a beta distribution, and
- some people regard a uniform $\text{Beta}(1,1)$ as uninformative,
- some might use a Jeffreys $\text{Beta}(0.5,0.5)$ prior, and
- some a Haldane improper $\text{Beta}(0,0)$ prior.
But since these are different, others might argue that simply choosing one would be informative.
