Markov Chains

Sequential Metropolis-within-Gibbs (SMWG)


Aspect
Description
Acceptance Rate The optimal acceptance rate is 44%, and is based on the univariate normality of each marginal posterior distribution. The observed acceptance rate may be suitable in the interval [15%,50%].
Applications This algorithm is applicable with state-space models (SSMs), including dynamic linear models (DLMs).
Difficulty This algorithm is relatively easy for a beginner when the proposal variance has been tuned with the SAMWG algorithm. Otherwise, it may be tedious for the user to tune the proposal variance.
Final Algorithm? Yes.
Proposal Componentwise.

The Sequential Metropolis-within-Gibbs (SMWG) algorithm is the non-adaptive version of the Sequential Adaptive Metropolis-within-Gibbs (SAMWG) algorithm, and is used for final sampling of state-space models (SSMs).


Bookmark this page