Methods for posterior inference of states and parameters.
# S3 method for nongaussian
run_mcmc(
model,
iter,
nsim,
output_type = "full",
mcmc_type = "da",
sampling_method = "psi",
burnin = floor(iter/2),
thin = 1,
gamma = 2/3,
target_acceptance = 0.234,
S,
end_adaptive_phase = TRUE,
local_approx = TRUE,
n_threads = 1,
seed = sample(.Machine$integer.max, size = 1),
max_iter = 100,
conv_tol = 1e-08,
...
)Model model.
Number of MCMC iterations.
Number of state samples per MCMC iteration.
Ignored if mcmc_type is "approx".
Either "full"
(default, returns posterior samples of states alpha and hyperparameters theta),
"theta" (for marginal posterior of theta),
or "summary" (return the mean and variance estimates of the states
and posterior samples of theta).
What MCMC algorithm to use? Possible choices are
"pm" for pseudo-marginal MCMC,
"da" for delayed acceptance version of PMCMC (default),
"approx" for approximate inference based on the Gaussian approximation of the model,
or one of the three importance sampling type weighting schemes:
"is3" for simple importance sampling (weight is computed for each MCMC iteration independently),
"is2" for jump chain importance sampling type weighting, or
"is1" for importance sampling type weighting where the number of particles used for
weight computations is proportional to the length of the jump chain block.
If "psi", \(\psi\)-auxiliary particle filter is used for state sampling
(default). If "spdk", non-sequential importance sampling based
on Gaussian approximation is used. If "bsf", bootstrap filter
is used.
Length of the burn-in period which is disregarded from the
results. Defaults to iter / 2.
Thinning rate. Defaults to 1. Increase for large models in
order to save memory. For IS-corrected methods, larger
value can also be statistically more effective.
Note: With output_type = "summary", the thinning does not affect the computations
of the summary statistics in case of pseudo-marginal methods.
Tuning parameter for the adaptation of RAM algorithm. Must be between 0 and 1 (not checked).
Target acceptance ratio for RAM. Defaults to 0.234.
Initial value for the lower triangular matrix of RAM algorithm, so that the covariance matrix of the Gaussian proposal distribution is \(SS'\). Note that for some parameters (currently the standard deviation and dispersion parameters of bsm_ng models) the sampling is done for transformed parameters with internal_theta = log(theta).
If TRUE (default), $S$ is held fixed after the burnin period.
If TRUE (default), Gaussian approximation needed for
importance sampling is performed at each iteration. If false, approximation is updated only
once at the start of the MCMC.
Number of threads for state simulation.
Seed for the random number generator.
Maximum number of iterations used in Gaussian approximation.
Tolerance parameter used in Gaussian approximation.
Ignored. set.seed(1) n <- 50 slope <- cumsum(c(0, rnorm(n - 1, sd = 0.001))) level <- cumsum(slope + c(0, rnorm(n - 1, sd = 0.2))) y <- rpois(n, exp(level)) poisson_model <- bsm_ng(y, sd_level = halfnormal(0.01, 1), sd_slope = halfnormal(0.01, 0.1), P1 = diag(c(10, 0.1)), distribution = "poisson") mcmc_is <- run_mcmc(poisson_model, iter = 1000, nsim = 10, mcmc_type = "is2") summary(mcmc_is, what = "theta", return_se = TRUE)