Gaussian Kernel Robust Regression (GKRReg)
Overview
Implements the Gaussian Kernel Robust Regression (GKRReg / GKRR) method proposed by De Carvalho, Lima Neto and Ferreira (2017) doi:10.1016/j.neucom.2016.12.035. The method re-weights observations iteratively using the Gaussian kernel so that poorly-fitted observations (outliers, leverage points) receive small weights, yielding resistance to Y-space outliers, X-space outliers and leverage points. Convergence is guaranteed by Propositions 4.1 and 4.2 of the original paper. Four estimators for the kernel width hyper-parameter are provided (S1: Caputo, S2: pairwise median, S3: residual variance, S4: corrected Akaike bandwidth), together with a data-driven selection rule. Inference is provided by a sandwich variance estimator with a family of heteroskedasticity-consistent finite-sample corrections adapted to the weighted leverage of the method, or by a pairs bootstrap that re-estimates the kernel width on each replicate (percentile, normal and bias-corrected accelerated intervals with p-values). Six real datasets from the robust regression literature are included to facilitate reproducible comparisons.
New features
gkrr_vcov()computes the sandwich covariance matrix with any member of the GKRReg-HC family of finite-sample corrections (HC0toHC4). The weighted leverage can be taken from the bread matrix (leverage = "bread", the quantity delivered by the leave-one-out expansion) or from the kernel-weighted hat matrix (leverage = "kernel", always in[0, 1]), andHC1/HC4can be calibrated by the effective sample sizesum(k_ii)(neff = "sum") or by Kish's effective sample size (neff = "kish").vcov()gains atypeargument dispatching togkrr_vcov(). The defaulttype = "HC0"reproduces the previous behaviour.gkrr_vcov_all()returns standard errors for every variant at once, together with diagnostics (both effective sample sizes, the number of observations with a negative bread weight, and the range of both leverages). Intended for simulation studies.
Show all changes in 0.5.0Hide the rest of the changes in 0.5.0
gkrr_jackknife()refits the model with each observation deleted in turn and compares the exact leave-one-out residual with the one-step approximation used byHC3andHC4.
Reproducibility
gkrr()andsigma2_s1()gain aseedargument.sigma_method = "s1"draws a random sub-sample and was previously **not reproducible**: repeated calls on the same data returned different estimates ofgamma^2and hence different coefficients. The caller's random number stream is left unchanged.
Documentation
- The vignette now covers the HC family inside its inference section, rather than as an appendix: the effective sample size, the weighted leverage and its admissible range, the effect of
kernel_update,gkrr_vcov_all()and the jackknife check, followed by a three-way comparison of HC0, the corrections and the bootstrap. - Man page titles harmonised to sentence case throughout, including the vignette title, which also appears in the package help index.
- The introductory vignette was renamed from
introductiontogkrreg, so thatvignette("gkrreg")is unambiguous (vignette("introduction")collided with vignettes of the same name in other packages). - The
summary()method for"gkrr_boot"objects now labels the printed quantiles of the bootstrap distribution. - The first author's name is no longer mangled in the package help page.
Authors@Rnow uses a Unicode escape, keeping DESCRIPTION pure ASCII, so the accented character survives documentation generation.
From NEWS.md in the release
Install
Health
- OK13 OK2026-08-05
- NOTE12 OK, 1 NOTE2026-08-01
- OK7 OK2026-06-18
Documentation
- Help pages
- 17 (1 internal)includes a package overview page
- Examples that run
- 100%11 run in a plain check
- Exported functions with examples
- 83%
- Documented parameters
- 100%
- Return-value docs
- 100%
- Help pages citing references
- 47%
Downloads
Repository
2 development-tooling and community-health practices detected across 2 families in the upstream repository
Checks run against github.com/marcelorpf/gkrreg on 2026-10-11.
GitHub Pages site last deployed 2026-06-17
Dependencies
Code & Tests
Datasets
People & History
Pick two releases to compare their code metrics.
- 0.5.0Latest2026-10-14 · current release · diff ↗ ··
added 3 exports
gkrr_jackknife, gkrr_vcov, gkrr_vcov_alllisted once in CRAN's queue
Seen in one reading of CRAN's incoming queue, at 2026-10-14 15:39 UTC, so it stayed less than the time between readings. - 0.4.02026-06-17
- RR 4.6.0 released · 2026-04-24
Package metadata
- First published
- 2026-06-17
- Total releases
- 2 / 1 yrs
- License
- GPL-3 OSI
- Minimum R
- ≥ 4.0.0
- Bundled data
- 3.8 KB / 6 files
- Download size
- 302 KB
Cite
Cite this package
Run in R for the authors' preferred citation:
citation("gkrreg")This package has no citation file, so citation() returns this entry built from DESCRIPTION.
Cite the R Observatory
For a number measured here: a download total, a coverage figure, an archival date.
From data release v2026-10-19, which the citation names so these numbers can be found later. More on citing and the projects behind them.