Sequential Bayesian updating — run_sequential

Refits the model on cumulative case subsets (cases 1-2, then 1-3, ...) with the same prior each refit, tracking how the treatment-effect estimate and its interval change as cases accrue. This is a cumulative-refit variant of sequential Bayesian updating; the posterior is not literally carried forward as the next prior.

Usage

run_sequential_update(
  data,
  case_var = "case",
  formula_str = NULL,
  family = "gaussian",
  iter = 2000,
  chains = 2,
  priors = NULL
)

Arguments

data: Data frame containing SCD data.
case_var: Name of the case variable.
formula_str: Model formula string; defaults to "outcome ~ phase + (1|case)".
family: Model family (default "gaussian").
iter: Number of sampling iterations.
chains: Number of MCMC chains.
priors: A brmsprior object applied to every refit. A supplied prior such as the analysis model's model$prior was scaled on all cases, so early refits then draw on outcome scale from cases not yet in the cumulative subset. NULL (default) fixes every prior class from the first two cases only, computed once so every refit shares the same prior: the package normal coefficient (b) and Student-t sd priors, plus the brms default Student-t Intercept and sigma priors evaluated on those two cases rather than recomputed on each refit's data; never brms flat priors.

Value

A list with tracking (data frame of cumulative estimates) and final_model.

Examples

if (interactive()) {
run_sequential_update(simulate_scd_data(n_cases = 5))
}