Prior Sensitivity Analysis for Bayesian Meta-Analysis

Assess how sensitive results are to different prior specifications

Usage

prior_sensitivity_analysis(
  yi,
  sei,
  study_labels = NULL,
  effect_measure = c("hr", "or", "rr", "rd", "md", "smd"),
  prior_scenarios = NULL,
  chains = 2,
  iter = 2000,
  seed = 42,
  ...
)

Arguments

yi

Numeric vector of effect estimates

sei

Numeric vector of standard errors

study_labels

Character vector of study names (optional)

effect_measure

Character. Effect type

prior_scenarios

List of prior specification scenarios. Each scenario should have: name: Character name for scenario prior_mu: List with mean and sd prior_tau: List with type and scale

chains

Integer. Number of MCMC chains per scenario. Default: 2

iter

Integer. Total iterations per chain. Default: 2000

seed

Integer. Random seed for reproducibility

...

Additional arguments passed to bayesian_meta_analysis()

Value

A list containing:

scenarios

List of bayesian_meta_result objects for each scenario

comparison

Data frame comparing estimates across scenarios

sensitivity_summary

Summary of how results change across priors

Examples

if (FALSE) { # \dontrun{
scenarios <- list(
  weak = list(
    prior_mu = list(mean = 0, sd = 10),
    prior_tau = list(type = "half_cauchy", scale = 0.5)
  ),
  informative = list(
    prior_mu = list(mean = 0, sd = 1),
    prior_tau = list(type = "half_cauchy", scale = 0.25)
  )
)

sensitivity <- prior_sensitivity_analysis(
  yi = yi,
  sei = sei,
  effect_measure = "hr",
  prior_scenarios = scenarios
)
print(sensitivity$comparison)
} # }