Estimate Resistance Profiles: Pathway 1 (Convex) or Pathway 2 (Bayesian)

Top-level dispatcher. Runs either the convex optimisation pathway (Pathway 1, aggregate surveillance data) or the Bayesian hierarchical multivariate probit pathway (Pathway 2, facility-level AST data).

Usage

estimate_resistance_profiles(
  data,
  method = c("convex", "bayesian"),
  pairwise = NULL,
  panel_map = NULL,
  class_cols = NULL,
  fixed_effects = NULL,
  random_effects = NULL,
  pathogen = NULL,
  pathogen_col = "pathogen",
  eligible_pairs = NULL,
  outcome_col = NULL,
  nonfatal_values = c("Discharged", "Survived", "Alive", "discharged", "survived",
    "alive"),
  panel_eligibility = list(),
  residual_structure = c("identity", "correlated"),
  estimand = "observed_stewardship_event_mix",
  prior_config = list(),
  sampler_config = list(),
  n_posterior_draws_for_profiles = 2000L
)

Arguments

data

For Pathway 1: marginals tibble from compute_marginals_from_data(). For Pathway 2: wide-format event-level tibble (one row per event; class columns 0/1/NA).

method

Character. "convex" or "bayesian".

pairwise

Tibble or NULL. Pathway 1 only.

panel_map

Named list or NULL. Pathway 1 only.

class_cols

Character vector. Pathway 2 only. Required.

fixed_effects

Character vector. Pathway 2 only. Required.

random_effects

Character vector (1-3 elements). Pathway 2 only. Required. Elements: hospital; [+patient]; [+admission].

pathogen

Character or NULL. Pathway 2 only. When supplied, filters data to a single pathogen before fitting.

pathogen_col

Character. Default "pathogen".

eligible_pairs

Tibble or NULL. Pathway 2 only.

outcome_col

Character or NULL. Pathway 2 only. Patient outcome column for R_ALL / R_NF split.

nonfatal_values

Character vector. Outcome values for R_NF cohort.

panel_eligibility

Named list. Eligibility thresholds; passed to fit_bayesian_multivariate_probit().

residual_structure

Character. Pathway 2. "identity" (default) or "correlated". See fit_bayesian_multivariate_probit().

estimand

Character. Estimand label. Default "observed_stewardship_event_mix".

prior_config

Named list. Pathway 2 only. Priors.

sampler_config

Named list. Pathway 2 only. Structured sampler settings: chains (4), iter_warmup (1000), iter_sampling (1000), seed (123), adapt_delta, max_treedepth, parallel_chains.

n_posterior_draws_for_profiles

Integer. Pathway 2. Number of posterior draws used for MVN simulation of profiles. Default 2000L.

Value

Named list: profiles, eligibility, diagnostics, fitted_models, config_used.

Details

For Pathway 2, pass pathogen to fit one pathogen at a time – the recommended workflow. Run this function once per pathogen and collect the results in the analysis repository.