Fit a Bayesian model using the brms package with default settings

Usage

fit_brms_model(
  ...,
  chains = 4,
  iterations = 2000,
  warmup = 1000,
  cores = chains,
  refresh = 500,
  backend = "rstan",
  file_refit = "on_change",
  file_compress = "xz",
  model_folder = "models/",
  sample_prior = FALSE,
  save_pars = NULL,
  adapt_delta = 0.95,
  max_treedepth = 10,
  seed = 667
)

Arguments

...: Arguments passed to brms::brm(), such as formula, data, family, priors, etc.
chains: Number of MCMC chains. Default is 4 (standard convention: enough for reliable Rhat/ESS convergence diagnostics on typical models, without the overhead of running far more chains than needed). Increase for models with tricky posteriors, not as a routine choice.
iterations: Number of POST-WARMUP iterations PER CHAIN (not divided by anything, not a total across chains). Default is 2000. Total post-warmup draws across all chains = iterations * chains.
warmup: Number of warmup iterations per chain. Default is 1000.
cores: Number of cores to use for parallel processing. Default is chains, i.e. one core per chain (fully parallel). Set lower only if the machine has fewer cores than chains requested.
refresh: Frequency of progress updates. Default is 500.
backend: Backend to use for fitting the model. Default is "rstan".
file_refit: Condition for refitting the model. Default is "on_change".
file_compress: Compression method for saving the model file. Default is "xz".
model_folder: Folder to save the fitted models. Default is "models/".
sample_prior: Logical. If TRUE, prior samples are drawn. If "only", only prior samples are drawn. Default is FALSE. FALSE
save_pars: Parameters to save. Default is NULL.
adapt_delta: Target acceptance rate for the NUTS sampler. Default is 0.95.
max_treedepth: Maximum treedepth for the NUTS sampler. Default is 10 (brms/Stan default). Increase (e.g. 12-15) if you see "maximum treedepth exceeded" warnings - this doesn't fix an underlying geometry problem, it just lets the sampler take more steps per iteration before giving up, which is often sufficient for models with awkward but not pathological posteriors (e.g. nonlinear/hinge models with a weakly identified parameter).
seed: Random seed for reproducibility. Default is 667.

Value

A fitted brms model object.