Robust Adaptive MCMC Sampler — robustAMCMC • caviarma

Implements a robust adaptive MCMC algorithm using the method of Vihola (2012) with Student-t proposals for improved robustness.

Usage

robustAMCMC(
  nsim,
  theta_seed,
  init_scale,
  acc_rate = 0,
  gamma = 2/3,
  adapt = FALSE,
  n_start = 0L,
  verbose = TRUE
)

Arguments

nsim: Total number of MCMC iterations
theta_seed: Initial parameter values
init_scale: Initial proposal scale vector (1/variance for each parameter)
acc_rate: Target acceptance rate (0 for no target, typically 0.234)
gamma: Adaptation decay rate, must be in (0.5, 1)
adapt: Enable adaptation (auto-enabled if acc_rate > 0)
n_start: Starting iteration for adaptation (0 = from beginning)
verbose: Print progress messages (default TRUE)

Value

Matrix of MCMC samples (rows are iterations, columns are parameters)

Details

Uses adaptive proposal covariance based on the acceptance rate. The proposal uses Student-t distribution with degrees of freedom equal to the parameter dimension for improved robustness.

References

Vihola, M. (2012). Robust adaptive Metropolis algorithm with coerced acceptance rate. Statistics and Computing, 22(5), 997-1008.