bnma

Bayesian random-effects network meta-analysis using the gemtc R package with JAGS (normal likelihood, identity link, 4 chains, 100,000 iterations after 5,000 adaptation steps, thinning of 10).

Parameters

• df: Data frame with columns study (chr), treatment (chr), mean (num), std.dev (num), sampleSize (int).
• greater_is_better: Logical. If TRUE, higher values rank better (e.g. accuracy). If FALSE, lower values rank better (e.g. error rate).

Return value

A named list with:

• ranking: Named numeric vector of SUCRA scores (0–1, higher = better rank).
• league: Named list with matrices md, lower, upper (posterior median and 95% credible interval bounds). Each is a treatment × treatment matrix.
• heterogeneity: Named list with sd, sd_lower, sd_upper (posterior median and 2.5th–97.5th percentile of the between-study standard deviation).
• convergence: Named list with rhat_max, ess_bulk_min, ess_tail_min across all model parameters.