Core 01: Overview of the glmbayesCore package

library(glmbayesCore)
#> Loading required package: MASS

Core 01: Overview of the glmbayesCore package

glmbayesCore is the compiled sampling engine behind glmbayes (formula interface: glmb(), lmb()) and the in-development lmebayes mixed-model package. End users normally install glmbayes; this vignette orients developers and advanced users to what lives in Core and how the engine vignettes are numbered.

Appendix-style Chapter-A* labels belong to glmbayes Part 5 only. Engine documentation here uses consecutive Core-01 … Core-11 topics (see table below). Formula-centric package overview and demos: vignette("Chapter-A01", package = "glmbayes").

Ecosystem

Package Role
glmbayesCore C++/OpenCL samplers, pfamily, Prior_Setup(), rglmb() / rlmb(), envelope pipeline
glmbayes User-facing GLM/LM Bayesian fits, S3 methods, tutorials (Chapters 00–18, appendices A01–A12)
lmebayes (dev) Mixed models: lmerb() / glmerb(). Setup layer in glmbayesCore (Part 4, Core-20); Gibbs engines staged in lmebayesCore

Primary R entry points

  • Prior_Setup(), Prior_Check() — default and custom priors (prior.R).
  • pfamily constructors (dNormal, dNormal_Gamma, …) — pair simfun and pfun for posterior and prior draws.
  • rglmb(), rlmb() — matrix-input i.i.d. posterior samplers used by glmbayes and block-Gibbs extensions.
  • glmbfamfunc(), EnvelopeBuild(), … — envelope construction and simulation pipeline (simulationpipeline.R, envelopeorchestrator.R).

Extension checklist (not yet a vignette): system.file("ADDING_PFAMILY.md", package = "glmbayesCore").

Vignette index by Part

The vignettes are grouped into Parts, as in glmbayes. Numbers are permanent identifiers, so a Part need not use consecutive numbers. Subsections use the Core-NN-S0M pattern.

Part 1: Orientation

Vignette Topic
Core-01 This overview

Part 2: Estimation and accept–reject sampling

Vignette Topic
Core-02 Estimation architecture; simfunctions and families
Core-03 Likelihood subgradient densities; .rNormalGLM_cpp pipeline
Core-04 Accept–reject: Gamma dispersion
Core-05 Accept–reject: Gaussian regression, independent normal–gamma
Core-09 Implementation companion (independent normal–gamma)
Core-10 Prior derivations for Prior_Setup()

Part 3: Envelopes, parallel CPU, and OpenCL

Vignette Topic
Core-06 Envelope-related functions (overview)
Core-07 Parallel sampling (RcppParallel)
Core-08 OpenCL-accelerated EnvelopeBuild
Core-11 OpenCL backend installation; large-model rglmb()

Part 4: Mixed-model design and priors (enabling lmerb())

Vignette Topic
Core-20 Overview: two-block structure, package-wide lineage chart, file map, helper naming
Core-20-S01 model_setup(), check_identifiability()
Core-20-S02 Prior_Setup_GLMM(): Block 1 / Block 2 prior calibration
Core-20-S03 pfamily_list(), dGamma_list()
Core-20-S04 Prior_SetupGroup(), normalize_group(), rNormal_reg_group(), rNormalGLM_reg_group()

Part 5: Two-block Gibbs engines (reserved; see below)

In R after install: browseVignettes("glmbayesCore"), vignette("Core-06", package = "glmbayesCore") for the envelope overview, or vignette("Core-20", package = "glmbayesCore") for the mixed-model setup.

Legacy map (glmbayes appendices → Core)

Historical glmbayes Part 5 appendix labels map to Core vignettes as follows:

glmbayes appendix glmbayesCore
A02 Core-02
A05 Core-03
A06 Core-04
A07 Core-05
A08 Core-06
A09 Core-07
A10 Core-08
A11 Core-09
A12 Core-10
(backend OpenCL install) Core-11

glmbayes-only appendices: A01 (package overview), A03 (methods vs glm), A04 (directional tail diagnostics). User OpenCL tutorial with glmb(): vignette("Chapter-16", package = "glmbayes").

Reserved numbers (future)

  • Core-12 — planned knitted guide from ADDING_PFAMILY.md (extension authors), Part 1.
  • Core-21–Core-24 — further Part 4 material as needed.
  • Core-25–Core-29 — Part 5, the two-block Gibbs engines (rlmerb(), rglmerb(), rLMM_reg*, rGLMM_reg*, two_block_*) once they move back from lmebayesCore (user tutorials stay in glmbayes Ch 17–18).

No placeholder .Rmd files are shipped for these slots; this section records intent only.