Control MCMC fitting process

Controls settings for the autofit process of the MCMC JAGS sampler (specifies termination criteria), and values for the convergence checks.

set_autofit_control(
  max_Rhat = 1.05,
  min_ESS = 500,
  max_error = NULL,
  max_SD_error = NULL,
  max_time = list(time = 60, unit = "mins"),
  sample_extend = 1000
)

set_convergence_checks(
  max_Rhat = 1.05,
  min_ESS = 500,
  max_error = NULL,
  max_SD_error = NULL,
  remove_failed = FALSE,
  balance_probability = TRUE
)

Arguments

max_Rhat: maximum value of the R-hat diagnostic. Defaults to 1.05.
min_ESS: minimum estimated sample size. Defaults to 500.
max_error: maximum value of the MCMC error. Defaults to NULL. Be aware that PEESE publication bias adjustment can have estimates on different scale than the rest of the output, resulting in relatively large max MCMC error.
max_SD_error: maximum value of the proportion of MCMC error of the estimated SD of the parameter. Defaults to NULL.
max_time: list with the time and unit specifying the maximum autofitting process per model. Passed to difftime function (possible units are "secs", "mins", "hours", "days", "weeks", "years"). Defaults to list(time = 60, unit = "mins").
sample_extend: number of samples to extend the fitting process if the criteria are not satisfied. Defaults to 1000.
remove_failed: whether models not satisfying the convergence checks should be removed from the inference. Defaults to FALSE - only a warning is raised.
balance_probability: whether prior model probability should be balanced across the combinations of models with the same H0/H1 for effect / heterogeneity / bias in the case of non-convergence. Defaults to TRUE.

Value

set_autofit_control returns a list of autofit control settings and set_convergence_checks returns a list of convergence checks settings.

Arguments

Value

See also