Crate plskit 

plskit — PLS regression with modern inference.

Single Rust crate; the same engine is consumed by the Python, R, and Julia wrappers (which call into this crate through their respective FFI layers).

Re-exports

pub use error::PlsKitError;

pub use error::PlsKitResult;

pub use find_k::FindKOptimalOpts;

pub use find_k::FindKOptimalOutput;

pub use find_k::FindKSequenceOpts;

pub use find_k::FindKSequenceOutput;

pub use find_k::FindKeepOptimalOpts;

pub use find_k::FindKeepOptimalOutput;

pub use find_k::Selector;

pub use find_k::pls1_find_k_optimal;

pub use find_k::pls1_find_k_sequence;

pub use find_k::spls1_find_k_optimal;

pub use find_k::spls1_find_k_sequence;

pub use find_k::spls1_find_keep_optimal;

pub use fit::FitOpts;

pub use fit::KSpec;

pub use fit::ParChoice;

pub use fit::Pls1Model;

pub use fit::pls1_fit;

pub use fit::spls1_fit;

pub use perm_alpha::PermAlphaOutput;

pub use perm_alpha::pls1_perm_alpha;

pub use perm_null::PermNullOpts;

pub use perm_null::PermNullOutput;

pub use perm_null::pls1_perm_null;

pub use pls3::Pls3FitOpts;

pub use pls3::Pls3Model;

pub use pls3::Pls3Scores;

pub use pls3::TransformWhich;

pub use pls3::pls3_fit;

pub use pls3::pls3_transform;

pub use pls3::plssvd_fit;

pub use pls3::plssvd_transform;

pub use pls3::spls3_fit;

pub use pls3_signal_test::Pls3ConfirmatoryTestOpts;

pub use pls3_signal_test::pls3_confirmatory_test;

pub use predict::pls1_predict;

pub use preprocess::PreprocessInput;

pub use preprocess::PreprocessResult;

pub use preprocess::preprocess;

pub use rotate::RotateOutput;

pub use rotate::RotationMethod;

pub use rotate::VarimaxArgs;

pub use rotate::rotate;

pub use rotation_stability::RotationStabilityMethod;

pub use rotation_stability::RotationStabilityOpts;

pub use rotation_stability::RotationStabilityOutput;

pub use rotation_stability::pls1_rotation_stability;

pub use signal_test::CIOpts;

pub use signal_test::ConfirmatoryArgs;

pub use signal_test::ConfirmatoryMethod;

pub use signal_test::ConfirmatoryTestInput;

pub use signal_test::ConfirmatoryTestOpts;

pub use signal_test::ConfirmatoryTestOutput;

pub use signal_test::SPLIT_NB_REROUTE_N_PERM;

pub use signal_test::SplitNbGateOutput;

pub use signal_test::pls1_confirmatory_test;

pub use signal_test::split_nb_gate;

Modules

error

Error types for plskit. Error type for plskit. All public functions return PlsKitResult<T>.

find_k

K selection (CV / BIC / sequence). K-selection: optimal (CV / BIC) and sequence (sequential test) methods.

fit

PLS1 model fitting (the NIPALS PLS1 model, computed by Improved Kernel PLS). PLS1 fit: the NIPALS PLS1 model, computed by Improved Kernel PLS (one shared component loop, pls1_component_loop, with the X backend; see pls1_kernel). Public entry points: pls1_fit, spls1_fit.

linalg

Low-level linear algebra helpers (standardize, row-subset, etc.). Linear-algebra and small-stat helpers shared across the core.

perm_alpha

PLS1 dual coefficients from a caller’s Gram matrix (pls1_perm_alpha), for callers that hold the standardized block themselves. PLS1 dual coefficients from a Gram matrix: pls1_perm_alpha.

perm_null

Permutation-null engine for signed per-voxel z statistics. Powers the TFCE / cluster-mass / max-stat downstream pipeline. Permutation-null engine for signed per-voxel z statistics on PLS1 β.

pls3

PLS3 / PLSSVD fit and transform (SVD on X’Y, no deflation). PLS3 / PLSSVD — symmetric X↔Y covariance analysis. Public entry points: pls3_fit (alias plssvd_fit) and pls3_transform (alias plssvd_transform).

pls3_signal_test

Confirmatory PLS3 / PLSSVD omnibus test at LV1. Confirmatory PLS3 / PLSSVD omnibus test at LV1. Two methods over one statistic: split_exact calibrates the held-out LV correlation by permutation, split_nb compares it against the NB t-reference instead.

predict

PLS1 prediction from a fitted model. Apply a fitted PLS1 model to new X.

preprocess

Public preprocess helper: validates/normalizes weights and standardizes (X, y). preprocess::preprocess_block is its multi-column-Y form; it is not re-exported at the crate root because the wrappers expose both forms under the one name preprocess, dispatching on the shape of Y. Public preprocess helper. Normalizes weights and standardizes (X, y).

resample

Resampling utilities (permutation indices, split-half indices). Resampling engines used by signal_test, sequential, and find_k. Crate-internal — public surface is the callers.

rng

Seeded RNG construction. RNG path for plskit. ChaCha8Rng + pre-computed child seeds.

rotate

Simple-structure rotation of PLS weights (varimax). Simple-structure rotation of PLS weights. Varimax is implemented via pairwise Kaiser sweeps; the RotationMethod enum is parameterized so promax / oblimin / geomin can land later without changing the surface.

rotation_stability

Rotation-stability diagnostic for PLS1. Rotation-stability diagnostic for PLS1. Standalone subsampling pass that asks “does varimax rotation make axes more replicable than they would have been on the unrotated NIPALS basis?” Output is a paired rotated-vs-unrotated variance ratio with paired-bootstrap CIs.

signal_test

Confirmatory PLS1 omnibus test at fixed K. Confirmatory PLS1 omnibus test at fixed K: five methods.

Structs

CIScalar

Resampling CI for a scalar functional, plus its SD. Marshalled to a frozen dataclass on the Python side; fields must not be reordered.

ConfirmatoryCI

Output of the confirmatory subsampling engine. Marshalled into ConfirmatoryCI at the wrapper layer.

Functions

version

Returns the CARGO_PKG_VERSION string (e.g. "0.0.1").

Type Aliases

Col

Re-export of faer’s owned and borrowed matrix/column types.

ColRef

Re-export of faer’s owned and borrowed matrix/column types.

Mat

Re-export of faer’s owned and borrowed matrix/column types.

MatRef

Re-export of faer’s owned and borrowed matrix/column types.