Search arXiv⌕ Search

arXiv · 2611.09484

Scalable Estimation and Inference for Varying Coefficient Expectile Regression with Streaming Data

Abstract

Nonparametric expectile regression methods face new challenges from streaming data, where retaining the full dataset for batch processing becomes increasingly impractical. In this paper, we develop a scalable nonparametric expectile regression framework for varying coefficient models. We estimate the coefficient functions online via functional stochastic gradient descent with Polyak--Ruppert averaging in a product reproducing kernel Hilbert space, requiring only gradient information and one pass over the data. For this online estimator, we establish non-asymptotic error bounds, prove its optimal nonparametric convergence rate under the proposed step-size schemes, and derive its Berry--Esseen bound for pointwise asymptotic normality. To perform online inference for the varying coefficient expectile regression models, we propose a multiplier bootstrap procedure that perturbs the stochastic gradients with independent random weights and recursively constructs bootstrap trajectories. We establish the bootstrap consistency of the online inference procedure, yielding asymptotically valid pointwise confidence intervals. Simulation studies and a real data application confirm the performance of the proposed methods.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Haojie Dong, Xinyuan Song, Xinghui Wang. 2026-11-06. Scalable Estimation and Inference for Varying Coefficient Expectile Regression with Streaming Data. https://arxiv.org/abs/2611.09484

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Optimal Forecast Reconciliation with Uncertainty Quantification

We propose to estimate the weight matrix used for forecast reconciliation as parameters in a general linear model in order to quantify its uncertainty. This implies that forecast reconciliation can be formulated as an orthogonal projection from the space of base-forecast errors into a coherent linear subspace. We use variance decomposition together with the Wishart distribution to derive the central estimator for the forecast-error covariance matrix. In addition, we prove that distance-reducing properties apply to the reconciled forecasts at all levels of the hierarchy as well as to the forecast-error covariance. A covariance matrix for the reconciliation weight matrix is derived, which leads to improved estimates of the forecast-error covariance matrix. We show how shrinkage can be introduced in the formulated model by imposing specific priors on the weight matrix and the forecast-error covariance matrix. The method is illustrated in a simulation study that shows consistent improvements in the log-score. Finally, standard errors for the weight matrix and the variance-separation formula are illustrated using a case study of forecasting electricity load in Sweden.

stat.ME↗

postcard: An R Package for Marginal Effect Estimation with or without Prognostic Score Adjustment

Covariate adjustment is a widely used technique in randomized clinical trials (RCTs) for improving the efficiency of treatment effect estimators. By adjusting for predictive baseline covariates, variance can be reduced, enhancing statistical precision and study power. Rosenblum and van der Laan [2010] use the framework of generalized linear models (GLMs) in a plug-in analysis to show efficiency gains using covariate adjustment for marginal effect estimation. Recently the use of prognostic scores as adjustment covariates has gained popularity. Schuler et al. [2022] introduce and validate the method for continuous endpoints using linear models. Building on this work Højbjerre-Frandsen et al. [2025] extends the method proposed by Schuler et al. [2022] to be used in combination with the GLM plug-in procedure [Rosenblum and van der Laan, 2010]. This method achieves semi-parametric efficiency under assumptions of additive treatment effects on the link scale. Additionally, Højbjerre-Frandsen et al. [2025] provide a formula for power approximation which is valid even under model misspecification, enabling realistic sample size estimation. This article introduces an R package, which implements the GLM plug-in method with or without PrOgnoSTic CovARiate aDjustment, postcard. The package has two core features: (1) estimating marginal effects and the variance hereof (with or without prognostic adjustment) and (2) approximating statistical power. Functionalities also include integration of the Discrete Super Learner for constructing prognostic scores and simulation capabilities for exploring the methods in practice. Through examples and simulations, we demonstrate postcard as a practical toolkit for statisticians.

stat.ME↗

Multi-layer dissolution exponential-family models for weighted signed networks

Understanding the structure of signed weighted networks is essential for analysing complex systems in which relationships between nodes vary both in sign and strength. Despite significant advances in statistical network analysis, there is still a lack of models that can account for both the sign and the strength of interactions in networks. We introduce a multi-layer dissolution exponential random graph modelling framework that enables systematic assessment of structural balance effects. We adopt a fully-probabilistic hierarchical approach that partially pools information across layers, with parameters estimated via an adaptive approximate exchange algorithm. A simulation study shows that the algorithm recovers the layer-specific parameters with small bias. We demonstrate the flexibility and explanatory power of the proposed methodology through an application to the gene interaction network associated with dilated cardiomyopathy, in which a multi-layer signed structure is constructed by progressively thresholding interaction confidence scores, with edge signs derived from pairwise co-expression. The analysis shows that positive co-expression edges embedded in positive triangles are more likely to persist at higher confidence levels, while negative edges whose endpoints share positive partners, which form unbalanced triangles, are preferentially dissolved as the confidence threshold increases.

stat.ME↗