Noise-resilient penalty operators based on statistical differentiation schemes

TL;DR

This paper introduces noise-resilient penalty operators for regression that operate directly on data grids, using statistically calibrated difference operators to handle discretely observed data with minimal smoothness assumptions.

Contribution

It proposes a novel class of penalized estimators based on statistical differentiation schemes that work directly on data grids, relaxing traditional smoothness constraints.

Findings

Estimators perform well in both smooth and irregular data settings.

Theoretical properties are established under Hellinger differentiability.

Simulation results show competitive performance.

Abstract

Penalized smoothing is a standard tool in regression analysis. Classical approaches often rely on basis or kernel expansions, which constrain the estimator to a fixed span and impose smoothness assumptions that may be restrictive for discretely observed data. We introduce a class of penalized estimators that operate directly on the data grid, denoising sampled trajectories under minimal smoothness assumptions by penalizing local roughness through statistically calibrated difference operators. Some distributional and asymptotic properties of sample-based contrast statistics associated with the resulting linear smoothers are established under Hellinger differentiability of the model, without requiring Fr\'echet differentiability in function space. Simulation results confirm that the proposed estimators perform competitively across both smooth and locally irregular settings.