Crate gamlss_formula

Expand description

Experimental optional typed formula/builder layer for compiling user model specifications into typed gamlss-core models.

gamlss-formula is a boundary layer: it reads user data through DataView, fits term metadata once, materializes predictor designs, and returns a BuiltModel artifact. Optimizers, fitting loops, diagnostics, dataframe ownership, and string formula parsing are intentionally outside this crate.

This crate is not the primary API of the project. The current primary path is the lower-level typed API in gamlss-core, gamlss-family, gamlss-spline, and gamlss-transform. gamlss-formula is an experimental convenience layer for curated high-level workflows and is not expected to cover every distribution, link, or parameterization available in the lower-level crates.

§Example

use gamlss_core::Objective;
use gamlss_formula::prelude::*;

let data = TestData {
    y: vec![0.0, 1.0, 2.0],
    x: vec![0.0, 1.0, 2.0],
};

let y = col::<f64>("y");
let x = col::<f64>("x");
let mut built = normal()
    .response(y)
    .mu(intercept() + linear(x))
    .sigma(intercept())
    .build(&data)?;

let theta = vec![0.0, 0.5, -0.2];
assert!(built.model_mut().value(&theta)?.is_finite());

Modules§

prelude: Common imports for the typed formula layer.

Structs§

BetaSpec: Typed beta model specification.
BuiltModel: Compiled formula artifact.
Category: Marker type for categorical columns.
Col: Typed reference to a named input column.
CyclicPSplineTerm: Cyclic P-spline term options.
FormulaPenalty: Formula-local segment penalty representation.
FormulaPredictorBlock: Formula predictor composed from a dense linear design, optional row offsets, and nonlinear monotone spline segments.
FourierTerm: Fourier term options.
GammaSpec: Typed gamma model specification.
InverseGaussianSpec: Typed inverse Gaussian model specification.
LogNormalSpec: Typed log-normal model specification.
ModelSchema: Schema for a built formula model.
ModelSpec: Entry point namespace for typed model specifications.
MonotoneTerm: Hard-monotone I-spline term options.
NormalSpec: Typed normal model specification.
PSplineTerm: Open-uniform P-spline term options.
ParameterTerms: Metadata for all terms belonging to one distribution parameter.
PredictionDesign: Reusable prediction design compiled from fitted term metadata.
ResponseSchema: Response metadata.
TensorPSplineTerm: Tensor-product P-spline term options.
TermExpr: Pre-data term expression for one parameter predictor.
WeibullSpec: Typed Weibull model specification.

Enums§

BoolCol: Boolean column storage returned by DataView.
CatCol: Categorical column storage returned by DataView.
FittedTerm: Fitted term metadata reusable for prediction.
FormulaError: Errors returned by the typed formula/builder layer.
NumericCol: Numeric column storage returned by DataView.
NumericResponse: Response storage used by compiled formula models.
TermSpec: Pre-data term specification.

Traits§

DataView: Read-only data access contract for the formula layer.

Functions§

beta: Creates a beta model spec.
col: Creates a typed column reference.
cyclic_pspline: Creates a cyclic P-spline term expression with default options.
factor: Creates a categorical factor term expression.
fourier: Creates a Fourier term expression with default options.
gamma: Creates a gamma model spec.
indicator: Creates a boolean indicator term expression.
interaction: Creates a numeric product interaction term expression.
intercept: Creates an intercept term expression.
inverse_gaussian: Creates an inverse Gaussian model spec.
linear: Creates a linear term expression.
log_normal: Creates a log-normal model spec.
monotone: Creates a hard-monotone I-spline term expression with default options.
no_intercept: Creates an explicit empty term expression without an intercept.
normal: Creates a normal model spec.
offset: Creates a numeric offset term expression.
pspline: Creates an open-uniform P-spline term expression with default options.
tensor_pspline: Creates a tensor-product P-spline term expression with default options.
weibull: Creates a Weibull model spec.

Type Aliases§

BetaBlocks: Blocks for a beta formula model.
BuiltBeta: Built beta formula model.
BuiltGamma: Built gamma formula model.
BuiltInverseGaussian: Built inverse Gaussian formula model.
BuiltLogNormal: Built log-normal formula model.
BuiltNormal: Built normal formula model.
BuiltWeibull: Built Weibull formula model.
CompiledBeta: Compiled beta formula model.
CompiledGamma: Compiled gamma formula model.
CompiledInverseGaussian: Compiled inverse Gaussian formula model.
CompiledLogNormal: Compiled log-normal formula model.
CompiledNormal: Compiled normal formula model.
CompiledWeibull: Compiled Weibull formula model.
FormulaBlock: Formula block type used by supported family specs.
GammaBlocks: Blocks for a gamma formula model.
InverseGaussianBlocks: Blocks for an inverse Gaussian formula model.
LogNormalBlocks: Blocks for a log-normal formula model.
NormalBlocks: Blocks for a normal formula model.
WeibullBlocks: Blocks for a Weibull formula model.