A self-contained MCMC sampler (no external backend). Runs multiple seeded chains, supports explicit starting values and run continuation, assesses convergence with several checkers, and returns the posterior draws with diagnostics.
Usage
morie_bayes_lm(
formula,
data,
prior_sd = 10,
chains = 4L,
iter = 2000L,
warmup = 1000L,
seed = 42L,
starting_values = NULL,
step = 0.1,
check_convergence = TRUE,
converge_threshold = 1.1,
stop_on_convergence = FALSE,
quiet = FALSE,
verbose = FALSE
)Arguments
- formula
A model formula.
- data
A data.frame.
- prior_sd
Prior standard deviation on the regression coefficients (a single value, or a vector of length
p). The meaning of this hyperparameter is the scale of the zero-mean Normal coefficient priors.- chains
Number of independent chains.
- iter
Iterations per chain (post-warmup).
- warmup
Warmup (burn-in) iterations.
- seed
Base RNG seed; each chain uses
seed + chain_indexso chains differ by default.- starting_values
Optional numeric vector (length
p + 1, the coefficients plus log-sigma) used as the starting state of every chain; or a list of such vectors, one per chain.- step
Metropolis proposal scale.
- check_convergence
Whether to compute convergence diagnostics.
- converge_threshold
R-hat threshold for declaring convergence.
- stop_on_convergence
Stop early once converged.
- quiet
Suppress warnings/progress.
- verbose
Emit per-chain progress messages.
Examples
d <- data.frame(x = rnorm(50)); d$y <- 1 + 2 * d$x + rnorm(50)
fit <- morie_bayes_lm(y ~ x, d, chains = 2, iter = 500, warmup = 200)