Type: Package
Title: Outrigger Regression
Version: 1.1.0
Maintainer: Elliot H. Young <ey244@cam.ac.uk>
Description: Performs outrigger local polynomial regression/ distributional adaptation, using a score-matching spline estimator of the conditional score function. Details of the method can be found in Young, Shah and Samworth (2026) <doi:10.48550/arXiv.2603.11282>.
License: GPL-3
Encoding: UTF-8
Imports: Rcpp, np, mgcv, RColorBrewer
LinkingTo: Rcpp
Suggests: testthat (≥ 3.0.0)
RoxygenNote: 7.2.3
NeedsCompilation: yes
Packaged: 2026-03-20 11:16:11 UTC; elliotyoung
Author: Elliot H. Young [cre], Rajen D. Shah [aut], Richard J. Samworth [aut]
Repository: CRAN
Date/Publication: 2026-03-24 10:10:15 UTC

Performs outrigger regression at a point x

Description

Performs outrigger regression at a point x

Usage

outrigger(
  formula,
  data,
  xtest,
  degree = 0,
  bandwidth = NULL,
  kernel = "epan",
  lambda = NULL,
  folds = 5,
  scoregranality = 1,
  score_df = 10,
  verbose = TRUE
)

Arguments

formula

formula (as in e.g. lm).

data

data frame.

xtest

datapoint(s) for local estimation.

degree

degree of polynomial.

bandwidth

bandwidth parameter. If NULL a rule-of-thumb bandwidth is selected by least-squares cross-validation for the standard local polynomial estimator.

kernel

kernel for local polyomial. Default is epanechnikov kernel.

lambda

orthogonalisation parameter. By default takes \lambda = 4(1 + 0.5 \log(n)).

folds

number of folds for cross-fitting. Default is 5.

scoregranality

the number of bins to split covariate-space |X-x|\leq\lambda h for conditional score estimation.

score_df

number of degrees of freedom for score matching splines used for conditional score estimation. Input can be a numeric value (e.g. 10 on larger datasets, 6 on smaller datasets) or "cross-validation" or "cross-validation-quick", in which case score-matching-CV will be performed (may be computationally costly).

verbose

suppresses messages of progress. Default is TRUE.

Value

If x is a single point, return a list containing:

prediction

The outrigger local polynomial estimator \hat{f}(x) at x.

fitted_vector

For the local linear outrigger estimator, the full fitted vector at x

standardlocpol_fitted_vector

The fitted vector for the standard local polynomial

score_plot_metadata

Data used for score_plotting to plot the fitted conditional score functions

If x is a vector of points, return a dataset with points x and associated outrigger fitted values.

Plots the fitted score functions learnt in an outrigger fit.

Description

Plots the fitted score functions learnt in an outrigger fit.

Usage

score_plots(fitted_outrig, plot_together = TRUE)

Arguments

fitted_outrig

a fitted object from outrigger

plot_together

a logical denoting whether all score function estimators (across all covariate bins) should be plotted toghether. Default is TRUE.

Value

No return value, called for plotting the estimated score function.