
128 Current Trends in Bayesian Methodology with Applications
developed for o nline infer e nce of static parameters (online Bayesian sequential
Monte Carlo or BSMC) by [14]). A dynamic linear model with Gaussian errors
is chosen since it is the easiest form of state space model, and would be a
good starting point to test our method. Average p e rformance is measured
across many replications of simulated data. To keep the computation-time
comparison fair, all methods were implemented in R [19]. T
INLA
is usually set
at so me value up till which INLA produces results extremely quickly. For
this example, it was fo und to be around 20, hence T
INLA
= 20.