Decorrelated Local Linear Estimator: Inference for Non-linear Effects in High-dimensional Additive Models

2025-12-31 · Source: JMLR · Field: Science & Research — Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

Zijian Guo, Wei Yuan, and Cunhui Zhang introduce a novel decorrelated local linear estimator (DLL) for inference in high-dimensional additive models, addressing the current lack of methods for confidence interval construction and hypothesis testing. The research, published in JMLR 27(27):1-79 in 2026, focuses on making inferences for function derivatives, particularly motivated by non-linear treatment effects. A key innovation is the development of decorrelation weights, which significantly reduce estimation error from nuisance functions. The proposed method facilitates the construction of confidence intervals and hypothesis testing for function derivatives. The authors validate their approach through extensive simulation studies and apply it to identify non-linear effects in motif regression problems. The DLL method is available as an R package on CRAN.

Key takeaway

For research scientists working with high-dimensional additive models and needing to infer non-linear relationships, the Decorrelated Local Linear (DLL) estimator provides a robust solution. You can now construct reliable confidence intervals and perform hypothesis testing on function derivatives, which was previously challenging. Consider integrating the DLL R package into your workflow to enhance the precision of your non-linear effect analyses, especially in applications like motif regression.

Key insights

A new decorrelated local linear estimator enables robust inference for non-linear effects in high-dimensional additive models.

Principles

• Decorrelation weights reduce nuisance function estimation error.
• Asymptotic normality supports confidence interval construction.

Method

The method involves constructing decorrelation weights for a local linear estimator to reduce error from nuisance function estimation, then applying it to establish asymptotic normality for function derivatives, enabling confidence interval construction and hypothesis testing.

In practice

• Construct confidence intervals for function derivatives.
• Perform hypothesis testing on non-linear effects.
• Analyze non-linear effects in motif regression.

Topics

• Decorrelated Local Linear Estimator
• High-dimensional Additive Models
• Non-linear Effects Inference
• Function Derivative
• Hypothesis Testing

Code references

• JmlrOrg/v27

Best for: AI Scientist, Research Scientist, Data Scientist

Related on AIssential

• Evidence Slopes and Effective Dimension in Singular Linear Models — Singular models require RLCT for accurate Bayesian evidence, as Laplace/BIC over-penalize by $(d/2-\lambda)\log n$.
• Kernel-Based Functional Balancing for Causal Inference with Compositional Treatments — A kernel-based functional balancing method and Augmented Weighted Estimator (AWE) enable √n-consistent causal inference for compositional treatments.
• Error Analysis for Deep ReLU Feedforward Density-Ratio Estimation with Bregman Divergence — BDD density-ratio estimators achieve minimax optimal error bounds and mitigate the curse of dimensionality.
• Claude Can’t Run Regression. A 200-Year-Old Theorem Proves It Never Will. — Regression estimates conditional expectations, requiring human judgment for true inference beyond pattern matching.
• Beyond Global Divergences: A Local-Mass Perspective on Bayesian Inference — A local-mass perspective on Bayesian inference reveals how global objectives overlook critical local distributional discrepancies.
• Generalized nonparametric regression in reproducing kernel Hilbert spaces: Consistency and rates of convergence — A new theory for RKHS M-estimation provides sharp convergence rates and explains how these estimators avoid the curse of dimensionality.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by JMLR.