Bayesian-like data analysis based on confidence distributions instead of prior distributions
What is it about?
This paper proposes using methods of imprecise probability for statistical inference on the basis of confidence distributions. A confidence distribution is a probability distribution on parameter space that, unlike a Bayesian posterior distribution, does not require a prior distribution.
Why is it important?
Bayesian statistics has the advantage of flexibly determining estimates and other decisions using the principle of maximum expected utility or, equivalently, minimum posterior expected loss. The main drawback is its need for a prior distribution. This paper proposes a simple method of overcoming that disadvantage using confidence intervals.
The following have contributed to this page: David R. Bickel
In partnership with: