Skip to contents

Changelog

Source: NEWS.md

bqmm 0.1.0

First release.

  • bqmm() — Bayesian mixed-effects quantile regression via the asymmetric Laplace likelihood and Stan, with an lme4-style formula interface (y ~ x + (1 + x | group)) supporting nested and crossed random effects.
  • cov = "diagonal" (default) or cov = "unstructured" — the latter adds LKJ-correlated random effects for a single grouping factor; VarCorr() reports the posterior-median correlation matrix.
  • Fitting one or several quantiles in a single call (a vector tau).
  • Post-hoc non-crossing rearrangement of fitted quantiles (Chernozhukov, Fernandez-Val and Galichon, 2010).
  • Yang, Wang and He (2016) posterior-variance correction for valid frequentist inference, exposed through vcov(object, adjusted = TRUE).
  • lme4/rstanarm-style S3 methods and integration with the posterior and bayesplot ecosystems.