| |
 |
| Paper: |
Computational Technology for Bayesian Inference |
| Monograph: |
6, Twenty Years of ADASS |
| Page: |
57 |
| Authors: |
|
| Abstract: |
One of the most important practical differences between the Bayesian approach to statistical inference and the more traditional frequentist approach is the nature of the sums and integrals required to implement each approach. In both approaches, the sampling distribution for the data (i.e., the likelihood function) plays a key role; but frequentist calculations integrate the sampling distribution over the sample space, whereas Bayesian calculations integrate it over hypothesis (parameter) space. The numerous advantages to working in parameter space come at a cost: the required integrals are difficult to calculate. I survey recent developments in computational technology for performing such integrals. |
|
|
|
|
|
