Module gamlss 

Multi-parameter GAMLSS family stack, organized into real concern modules.

The historical single-file module was split textually with include!; this is the genuine module decomposition. The fitting math is one tightly-coupled namespace (families share layouts, the location-scale joint-ψ machinery, and the numeric foundation), so the concern submodules each open with use super::*; and this parent re-exports their cross-concern items flat via pub(crate) use <module>::*. That keeps every intra-stack reference resolving without a web of explicit per-symbol imports, while each file now holds a single concern:

• errors — GamlssError, module-wide numeric constants, the low-level row-map / design-resolution / joint-ψ-direction helpers, and the ParameterLink link-identifier enum (the shared foundation every family builds on).
• builders — block/penalty construction, term specs, fit-result types, the LocationScaleFamilyBuilder strategy + its four builders, and the public fit_* entry points (construction & orchestration).
• gaussian — the Gaussian location-scale family (+ its wiggle variant), the shared location-scale joint-ψ trait/workspace, the Gaussian joint-Hessian assembly, the matrix-free RowCoeffOperator, and the Poisson/Gamma log-link families.
• binomial — the binomial location-scale numeric kernel (q-algebra, towers, directional coefficients) and the binomial location-scale / wiggle / mean-wiggle families plus their exact-Newton Hessian workspaces.

Pre-existing leaf concerns keep their own modules: dispersion_family, binomial_q_derivs, binomial_q_coeffs, validation, weighted_design_products, row_linalg, and joint_packing.

Structs

BinomialLocationScaleAloRowInput

Complete saved-row state for exact binomial location-scale ALO replay.

BinomialLocationScaleFamily

BinomialLocationScaleFitResult

BinomialLocationScaleTermSpec

BinomialLocationScaleWiggleFamily

BinomialLocationScaleWiggleTermSpec

BinomialMeanWiggleFamily

BinomialMeanWiggleTermFitResult

BlockwiseTermFitResult

BlockwiseTermWiggleFitResult

DispersionAloRowGeometry

Exact row-local geometry consumed by saved-model case deletion.

DispersionGlmLocationScaleTermSpec

Term spec consumed by fit_dispersion_glm_location_scale_terms; mirrors GaussianLocationScaleTermSpec with the dispersion channel in place of the Gaussian log-σ channel.

FamilyMetadata

GammaLogFamily

Built-in Gamma log-link family (single parameter block, fixed shape).

GaussianJointRowProgram

Production gam_math::jet_tower::RowProgram for the normalized Gaussian location-scale row NLL.

GaussianJointRowScalars

GaussianLocationScaleAloRowInput

Complete saved-row state for exact Gaussian location-scale ALO replay.

GaussianLocationScaleFamily

GaussianLocationScaleFitResult

GaussianLocationScaleTermSpec

GaussianLocationScaleWiggleFamily

GaussianLocationScaleWiggleTermSpec

LocationScaleAloRowGeometry

Exact row-local saved-model geometry in affine likelihood coordinates.

PoissonLogFamily

Enums

DispersionFamilyKind

The genuine-dispersion mean family whose precision (overdispersion) channel can carry a second noise_formula linear predictor (issue #913).

GamlssError

Typed errors surfaced from this module’s helpers and family implementations. The Display impl writes the carried reason verbatim, so callers that historically returned Result<_, String> keep their user-visible text byte-for-byte identical after coercion via the From<GamlssError> for String impl below.

ParameterLink

Link identifiers for distribution parameters in multi-parameter GAMLSS families.

Constants

FAMILY_BETA_LOCATION_SCALE

FAMILY_GAMMA_LOCATION_SCALE

FAMILY_NEGBIN_LOCATION_SCALE

FAMILY_TWEEDIE_LOCATION_SCALE

Functions

binomial_location_scale_alo_row_geometry

Replay one binomial threshold-scale row in affine saved coordinates.

dispersion_alo_row_geometry

Replay the exact fitted row likelihood in its two affine predictor coordinates for saved-model ALO.

fit_dispersion_glm_location_scale_terms

Fit a dispersion-channel GAMLSS location-scale model (#913). All four genuine-dispersion mean families share this single entry; the per-family likelihood lives in dispersion_row_kernel.

gaussian_location_scale_alo_row_geometry

Replay one Gaussian location-scale row in the raw saved coefficient frame.