Bootstrap Inference with Debiased Nonparametric Estimators

Implements debiased kernel density and local-linear regression estimators with empirical-bootstrap simultaneous confidence bands, as proposed by Cheng and Chen (2019) . Also provides bandwidth selectors and grid-based confidence sets for density level sets and inverse regression.

debiasedInference

R implementation of the procedures in Cheng and Chen, “Nonparametric Inference via Bootstrapping the Debiased Estimator”.

If you use these methods or this software, please cite the paper (Electronic Journal of Statistics 13(1), 2019; doi:10.1214/19-EJS1575).

library(debiasedInference)

set.seed(2026)
x <- rnorm(300)
fit <- debiased_kde(x)
band <- kde_confidence_band(x, n_boot = 499, random_state = 2026)

The result objects are ordinary lists with classes di_estimate, di_band, and di_set, so their numerical components are directly accessible.