GitHub

CPPLS.jl

CPPLS provides a pure-Julia implementation of Canonical Powered Partial Least Squares (CPPLS) for both regression and discriminant analysis. The package is intended for high-dimensional and collinear predictor settings in which latent-variable models are used for supervised dimension reduction, interpretation, and prediction.

Installation

The package is not registered, so install it directly from GitHub:

julia> ]
pkg> add https://github.com/oniehuis/CPPLS.jl

After the installation finishes you can load it in the Julia REPL with:

julia> using CPPLS

Scope

The package supports supervised latent-variable modelling with either matrix-valued responses or class labels. Fitting is controlled through CPPLSModel, which separates model configuration from the data passed to fit. The resulting fitted models provide prediction, class assignment, latent projections, regression coefficients, fitted values, residuals, and labels or metadata retained from model fitting.

CPPLS also includes helper functionality for common preprocessing and encoding tasks, including one-hot encoding of class labels and utility functions used in chemometric and multivariate workflows.

Validation and Inference

The cross-validation interface is designed around explicit evaluation callbacks, so the same machinery can be used for either discriminant analysis or regression. For most users, the preferred entry points are the higher-level wrappers cvda, permda, cvreg, and permreg, which apply standard defaults for common workflows. Lower-level callback bundles are also available through CPPLS.cv_classification() and CPPLS.cv_regression() for direct use with nestedcv and nestedcvperm. For classification workflows, outlierscan performs repeated outer-fold evaluation to quantify how often individual samples are misclassified.

Permutation-based significance assessment is available through nestedcvperm together with pvalue, allowing empirical evaluation of whether observed predictive performance exceeds what would be expected under response permutation.

Usage

For fitting and model specification, see CPPLS/fit.md. Longer worked workflows for discriminant analysis and regression are collected in CPPLS/fit_examples.md. Prediction, projection, and class-assignment methods are described in CPPLS/predict.md. Cross-validation, permutation testing, and outlier scanning are documented in CPPLS/crossvalidation.md. The main data structures and fitted-model accessors are summarized in CPPLS/types.md, while utility functionality is described under the Utils section of the manual.

Disclaimer

CPPLS is provided "as is," without warranty of any kind. Users are responsible for independently validating all outputs and interpretations and for determining suitability for their specific applications. The authors and contributors disclaim any liability for errors, omissions, or any consequences arising from use of the software, including use in regulated, clinical, or safety-critical contexts.

References

  • Indahl UG, Liland KH, Naes T (2009) Canonical partial least squares — a unified PLS approach to classification and regression problems. Journal of Chemometrics 23: 495-504. https://doi.org/10.1002/cem.1243.
  • Liland KH, Indahl UG (2009): Powered partial least squares discriminant analysis. Journal of Chemometrics 23: 7-18. https://doi.org/10.1002/cem.1186.
  • Smit S, van Breemen MJ, Hoefsloot HCJ, Smilde AK, Aerts JMFG, de Koster CG (2007): Assessing the statistical validity of proteomics based biomarkers. Analytica Chimica Acta 592: 210-217. https://doi.org/10.1016/j.aca.2007.04.043.