Point Estimation in Bayesian Statistics
- tags
- Bayesian Statistics
Suppose we have a posterior distribution over parameters
To do so, we may select a summary feature of
For symmetric posterior densities, the mean and median are equal. If the posterior is also unimodal, then all 3 coincide.
For asymmetric posteriors, the median is often preferred. The mode only considers the value corresponding to the maximum value of the density, while the mean may give too much weight to extreme outliers