Skip to contents

Run RoBMA Models Across Multiple Prior Specifications

Source: R/robma_sensitivity.R
robma_sensitivity.Rd

Pre-computes RoBMA models for each prior specification provided. These fits are stored and can later be attached to a bayesma object for use in sensitivity_plot(). This separation allows the computationally expensive RoBMA fitting to be done once and reused.

Usage

robma_sensitivity(
  data,
  priors,
  robma_template = NULL,
  robma_args = NULL,
  parallel = FALSE,
  workers = NULL,
  seed = TRUE,
  .progress = TRUE
)

Arguments

data

A data frame containing the study data.

priors

A named list of prior specifications. Each element must be a list with at least mu_prior and (optionally) tau_prior, and may include name used for display. Example: list( prior_1 = list(name = "Vague", mu_prior = normal(0, 10), tau_prior = half_cauchy(0, 1)), prior_2 = list(name = "Regularising", mu_prior = normal(0, 1), tau_prior = half_cauchy(0, 0.5)) )

robma_template

Either:

  • A fitted RoBMA object (class bayesma_robma) with stored $meta$call_args to use as a template for refitting across priors, OR

  • NULL (default), in which case robma_args must be provided

robma_args

A list of arguments to pass to robma() when robma_template is NULL. Must include all required arguments except priors_effect and priors_heterogeneity, which will be set from the priors argument.

parallel

Logical. If TRUE, uses parallel processing for refits.

workers

Optional integer. Number of parallel workers.

seed

Logical. If TRUE (default), uses parallel-safe seeding.

.progress

Logical. If TRUE, displays a progress bar.

Value

A bayesma_robma_sensitivity object: a named list of RoBMA fits keyed by prior IDs, with additional metadata in $meta.

Details

The returned object can be attached to a bayesma model using attach_robma_sensitivity() or passed directly to sensitivity_plot().

Examples

if (FALSE) { # \dontrun{
# Define priors for sensitivity analysis
priors <- list(
  vague = list(
    name = "Vague",
    mu_prior = normal(0, 10),
    tau_prior = half_cauchy(0, 1)
  ),
  informative = list(
    name = "Informative",
    mu_prior = normal(0, 0.5),
    tau_prior = half_cauchy(0, 0.3)
  )
)

# Option 1: Using a template RoBMA fit
robma_sens <- robma_sensitivity(
  data = my_data,
  priors = priors,
  robma_template = my_robma_fit,
  parallel = TRUE
)

# Option 2: Specifying robma arguments directly
robma_sens <- robma_sensitivity(
  data = my_data,
  priors = priors,
  robma_args = list(
    studyvar = "study_id",
    event_ctrl = "events_c",
    event_int = "events_i",
    n_ctrl = "n_c",
    n_int = "n_i",
    likelihood = "binomial"
  ),
  parallel = TRUE
)

# Attach to bayesma model
model <- attach_robma_sensitivity(model, robma_sens)

# Now sensitivity_plot can use the pre-computed RoBMA fits
sensitivity_plot(model, data, priors, estimand = "OR", incl_robma = TRUE)
} # }