Gibbs sampling in Analytica
From Opasnet
| Moderator:Nobody (see all) Click here to sign up. |
|
|
| Upload data
|
Scope
What is a good way to perform Gibbs sampling in Analytica such that
- it is easy to operationalise Bayesian updating within a model.
Definition
Input
The input should contain data and likelihood and prior distributions. This should be able to build for several model nodes, as they may serve as data for each other.
Output
The output should contain a posterior distribution.
Rationale
A tricky question is, what to do with probabilistic nodes, as in Analytica the default is that nodes are functional nodes. Can the definition of a node be described probabilistically?
Result
Procedure
Gibbs sampler should be some kind of function. It should have a variable list and a likelihood list as input parameters.
