Research map / Learning theory and algorithms / Sample complexity
Estimation and Inference for High Dimensional Generalized Linear Models: A Splitting and Smoothing Approach
Zhe Fei, Yi 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 |
|---|---|---|---|
| swap LASSO for Bayesian LASSO as the selection step (ablation) | SSGLM · Simulation Example 2 (Poisson) | average coverage probability: 0.93 -> 0.94 | No clear effect |
| swap LASSO for Bayesian LASSO as the selection step (ablation) | SSGLM · Simulation Example 2 (Poisson) | average model size: 23.12 -> 7.98 | Mixed |
| swap LASSO for Bayesian LASSO as the selection step (ablation) | SSGLM · Simulation Example 2 (Poisson) | average MSE: 0.105 -> 0.102 | No clear effect |
| use equal split proportion q=0.5 rather than other q (ablation) | SSGLM · Example 1 (linear, n=500, p=1000) | average MSE: minimum MSE at q=0.5 | Improved |
| increase split proportion q beyond 0.5 (ablation) | SSGLM · Example 1 (linear, n=500, p=1000) | average MSE: MSE less sensitive to q for larger q | No clear effect |
| use bootstrap-smoothed variance estimator V^B instead of plain infinitesimal jackknife V (ablation) | SSGLM · Example 1 (B=100 to 2000) | variance estimate convergence: V^B small bias even with small B; V converges slower | Improved |
| use SSGLM instead of de-biased LASSO (Lasso-pro) | Lasso-pro · Example 5 (rho=0.75) | power for beta10: 0.690 -> 0.863 | Improved |
| use SSGLM instead of de-biased LASSO (Lasso-pro) | Lasso-pro · Example 5 (rho=0.25) | computing time (secs): 74.7 -> 17.7 | Improved |
| use SSGLM instead of de-correlated score test | Dscore · Example 5 (rho=0.75) | type I error: 0.400 -> 0.060 | Improved |
| use SSGLM instead of de-correlated score test | Dscore · Example 5 (rho=0.6) | power for beta30: 0.647 -> 0.880 | Improved |
| add significant SSGLM interactions to demographic baseline model | Baseline logistic model · BLCSC lung cancer data | R^2: 0.0938 -> 0.1168 | Improved |
| add significant de-sparsified LASSO interactions to baseline model | Baseline logistic model · BLCSC lung cancer data | R^2: 0.0938 -> 0.1018 | Improved |
Closest papers by meaning
- High-Dimensional Inference for Generalized Linear Models with Hidden Confounding (JMLR 2023 · similarity 0.87)
- Targeted Fused Ridge Estimation of Inverse Covariance Matrices from Multiple High-Dimensional Data Classes (JMLR 2020 · similarity 0.87)
- Inference for the Case Probability in High-dimensional Logistic Regression (JMLR 2021 · similarity 0.87)
- No Double Descent in Principal Component Regression: A High-Dimensional Analysis (ICML 2024 · similarity 0.85)
- Risk and cross validation in ridge regression with correlated samples (ICML 2025 · similarity 0.84)
- Robust group and simultaneous inferences for high-dimensional single index model (NeurIPS 2024 · similarity 0.84)
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.
