gtmeta 0.1.0

gtmeta 0.1.0

First release.

• Study constructors for univariate (igmi_studies()), multivariate
  (igmi_gaussian()), diagnostic test accuracy (igmi_dta_studies()), and
  prediction-model performance study objects.
• Bures-Wasserstein barycenter pooling (bw_barycenter()) with the exact
  fixed-point iteration, descent/trace invariants as runtime checks, and
  reduction to the inverse-variance fixed-effect estimate in the scalar
  precision-weighted case.
• Fisher-Rao distances and means with the Karcher-ball admissibility check.
• Frechet-variance heterogeneity measures that recover classical Q, I2,
  and the DerSimonian-Laird moment estimator exactly in the scalar limit.
• Exact location inference (chi-square ellipsoids), Frechet-scatter
  (HKSJ-type) inference, and bootstrap-over-studies defaults.
• Robust Wasserstein-Fisher-Rao pooling (wfr_pool()), a redescending
  Andrews-sine M-estimator with transport cutoff pi * delta, plus an
  optional Python/POT bridge for general WFR distances.
• Classical fixed-effect, random-effects, and UWLS benchmarks with AIC/BIC
  and leave-one-study-out predictive scoring.
• Simulation generators, evaluation metrics, and adapters for Cochrane-style
  extracted estimates.
• Theorem-anchored test suite: every estimator is tested against a numbered
  result in the accompanying mathematical notes.