Research map / Learning theory and algorithms / Sample complexity
Statistical Inference for High-dimensional Partially Linear Models via Debiased Rank Lasso
Songshan Yang, Delin Zhao, Runze Li
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 |
|---|---|---|---|
| use data-split rank Lasso instead of two-step projection Lasso under Cauchy errors | partially linear model estimator · Simulation Ex.1, Cauchy, N=450 | l1 error: 5.185 -> 1.907 | Improved |
| use data-split rank Lasso instead of two-step Lasso under t(2) errors | partially linear model estimator · Simulation Ex.1, t(2), N=450 | l1 error: 1.710 -> 1.051 | Improved |
| use data-split rank Lasso instead of two-step Lasso under log-normal errors | partially linear model estimator · Simulation Ex.1, LogN, N=450 | l1 error: 0.928 -> 0.610 | Improved |
| use data-split rank Lasso instead of two-step Lasso under normal errors | partially linear model estimator · Simulation Ex.1, N(0,2^2), N=450 | l1 error: 0.926 -> 0.976 | No clear effect |
| use data-split rank Lasso to reduce false positives | partially linear model estimator · Simulation Ex.1, N(0,2^2), N=450 | false positives: 11.46 -> 0.200 | Improved |
| use data-split rank Lasso to avoid false negatives under Cauchy errors | partially linear model estimator · Simulation Ex.1, Cauchy, N=450 | false negatives: 1.680 -> 0.010 | Improved |
| use ANOPE subsampled loss instead of full pairwise rank loss (ablation) | DSRLasso · Simulation Ex.1, Cauchy, N=450 | l1 error: 1.907 -> 1.943 | No clear effect |
| use debiased DSRLasso instead of debiased TSLasso for CI length under Cauchy errors | debiased estimator · Simulation Ex.2, Cauchy, N=450 | average CI length (active set): 12.90 -> 1.104 | Improved |
| use debiased DSRLasso instead of debiased TSLasso for power under Cauchy errors | debiased estimator · Simulation Ex.2, Cauchy, N=450 | power (ARPB): 0.438 -> 0.992 | Improved |
| use debiased DSRLasso instead of debiased TSLasso for simultaneous CI length under Cauchy errors | debiased estimator · Simulation Ex.3 NST, Cauchy, N=450 | average simultaneous CI length: 26.01 -> 2.868 | Improved |
| use debiased DSRLasso instead of debiased TSLasso for simultaneous test power under Cauchy errors | debiased estimator · Simulation Ex.3 NST, Cauchy, N=450 | rejection probability: 0.320 -> 0.985 | Improved |
| use debiased DSRLasso instead of debiased TSLasso for coverage under normal errors | debiased estimator · Simulation Ex.2, N(0,2^2), N=300 | coverage of beta_1 (CP1): 0.870 -> 0.915 | Improved |
Closest papers by meaning
- Robust Inference for High-Dimensional Linear Models via Residual Randomization (ICML 2021 · similarity 0.88)
- Scalable Resampling in Massive Generalized Linear Models via Subsampled Residual Bootstrap (JMLR 2024 · similarity 0.86)
- Communication-efficient Sparse Regression (JMLR 2017 · similarity 0.86)
- L1-Regularized Least Squares for Support Recovery of High Dimensional Single Index Models with Gaussian Designs (JMLR 2016 · similarity 0.86)
- Rank-based Lasso - efficient methods for high-dimensional robust model selection (JMLR 2020 · similarity 0.85)
- Inference In High-dimensional Single-Index Models Under Symmetric Designs (JMLR 2021 · similarity 0.85)
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.
