Skip to contents

bhfvar 0.3.0

Initial Release

This is the initial release of the bhfvar package, implementing the Bayesian Hybrid Framework for variance decomposition in complex surveys with post-hoc domains.

Key Features

  • Bayesian Pseudo-Likelihood: Design-consistent inference using survey weights in a Bayesian framework

  • Hybrid GLMM: Simultaneous modeling of substantive domain effects (states) and nuisance design effects (PSUs)

  • Dual Estimand Framework:

    • Estimand A (Policy): Logit-scale variance decomposition
    • Estimand B (Descriptive): Probability-scale variance decomposition
    • Estimand A* (Policy Adjusted): De-attenuated variance decomposition
  • De-attenuation: Both implicit (Bayesian shrinkage) and explicit (method-of-moments) correction for finite-sample variance inflation

  • Comprehensive Output:

Main Functions

Documentation

  • Comprehensive vignettes covering:
    • Introduction to the package
    • Quick start guide
    • Methodology details
    • Dual estimand framework
    • Diagnostics and model checking
    • API reference
  • pkgdown website configuration for online documentation

Dependencies

  • R >= 4.0.0
  • rstan >= 2.26.0
  • posterior >= 1.0.0