Multi-Site Federated Meta-Analysis Aggregator — federated_meta

Computes pooled effect estimates across federated DuckDB nodes using the DerSimonian-Laird random-effects or inverse-variance fixed-effect method, without sharing patient-level data. Computes Cochran's Q, tau^2, and Higgins' I^2.

Usage

federated_meta_analysis(
  estimates = NULL,
  std_errors = NULL,
  variances = NULL,
  sites = NULL,
  site_summaries = NULL,
  con = NULL,
  query = NULL,
  model = "random",
  effect_type = "beta",
  alpha = 0.05
)

Arguments

estimates: Vector of study-level parameter estimates (e.g. log OR, log HR, beta).
std_errors: Vector of study standard errors.
variances: Optional vector of study variances (used if std_errors is NULL).
sites: Optional character vector of site identifiers.
site_summaries: Optional data.frame with columns site, estimate, std_error.
con: Optional active federated DuckDB connection.
query: Optional SQL query to evaluate across site tables or v_* views.
model: Model type: "random" (DerSimonian-Laird, default) or "fixed".
effect_type: Effect metric: "beta", "log_or", "log_hr", or "mean_diff".
alpha: Significance level for confidence intervals (default 0.05).

Value

An object of class federated_meta_analysis containing pooled estimates and statistics.