proposal_covariance
Defines the technique used to generate the MCMC proposal covariance.
Topics
bayesian_calibration
Specification
Alias: None
Arguments: None
Child Keywords:
Required/Optional |
Description of Group |
Dakota Keyword |
Dakota Keyword Description |
---|---|---|---|
Required (Choose One) |
Proposal Covariance Source |
Uses the covariance of the prior distributions to define the MCMC proposal covariance. |
|
Use derivatives to inform the MCMC proposal covariance. |
|||
Specifies matrix values to use as the MCMC proposal covariance. |
|||
Uses a file to import a user-specified MCMC proposal covariance. |
Description
The proposal covariance is used to define a multivariate normal (MVN) jumping distribution used to create new points within a Markov chain. That is, a new point in the chain is determined by sampling within a MVN probability density with prescribed covariance that is centered at the current chain point. The accuracy of the proposal covariance has a significant effect on rejection rates and the efficiency of chain mixing.
Default Behavior
The default proposal covariance is prior
when no emulator is
present; derivatives
when an emulator is present.
Expected Output
The effect of the proposal covariance is reflected in the MCMC chain
values and the rejection rates, which can be seen in the diagnostic
outputs from the QUESO solver within the QuesoDiagnostics
directory.
Usage Tips
When derivative information is available inexpensively (e.g., from an emulator model), the derived-based proposal covariance forms a more accurate proposal distribution, resulting in lower rejection rates and faster chain mixing.