Research map / Learning theory and algorithms / Kernels and regression theory
Can We Trust the Bootstrap in High-dimensions? The Case of Linear Models
Noureddine El Karoui, Elizabeth Purdom
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase p/n from small to 0.5 | residual bootstrap CIs · Least squares, n=500 | 95% CI error rate: 10-15% at p/n = 0.3, approximately 20% at p/n = 0.5 | Hurt |
| increase p/n for pairs bootstrap | pairs bootstrap CIs · n=500, L1/Huber/L2 | 95% CI error rate: error rate drops far below 5%, essentially zero at p/n = 0.5 | Hurt |
| use L1 loss with pairs bootstrap at p/n=0.1 | pairs bootstrap CIs · L1 loss, n=500 | 95% CI error rate: dropping to less than 0.01 | Hurt |
| bootstrap from standardized predicted errors instead of raw residuals (ablation) | residual bootstrap · L1 and Huber, double exponential errors | 95% CI error rate | Improved |
| use weighted bootstrap with weights from Equation (6) instead of Poisson(1) | pairs bootstrap · L2, n=500, Gaussian design | 95% CI error rate: 0.051, 0.06, 0.061, 0.057 for kappa .1, .2, .3, .5 | Improved |
| slightly misspecify alpha (0.95 instead of 0.92) (ablation) | weighted pairs bootstrap · L2, kappa=0.5 | bootstrap variance: roughly 30% larger than it should be | Hurt |
| rescale jackknife variance by 1 - p/n (ablation) | jackknife · L2, Gaussian design | 95% CI error rate | Improved |
| apply the 1 - p/n jackknife rescaling to elliptical design (ablation) | jackknife · L2, elliptical design | 95% CI error rate | No clear effect |
| apply the jackknife correction with L1 loss (ablation) | jackknife · L1, n=1000 | 95% CI error rate | Hurt |
Closest papers by meaning
- Bootstrap in High Dimension with Low Computation (ICML 2023 · similarity 0.92)
- Distributed Bootstrap for Simultaneous Inference Under High Dimensionality (JMLR 2022 · similarity 0.86)
- Small Resamples, Sharp Guarantees: Convergence Rates for Resampled Studentized Quantile Estimators (NeurIPS 2025 · similarity 0.86)
- Asymptotics of the Bootstrap via Stability with Applications to Inference with Model Selection (NeurIPS 2021 · similarity 0.86)
- Risk and cross validation in ridge regression with correlated samples (ICML 2025 · similarity 0.84)
- High-Dimensional Analysis for Generalized Nonlinear Regression: From Asymptotics to Algorithm (AAAI 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
