Robust spatial scalar-on-function regression: A Fisher-consistent redescending M-estimation approach
Muge Mutis · 2026
Read Robust spatial scalar-on-function regression: A Fisher-consistent redescending M-estimation approach by Muge Mutis (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-00470.txt.
Description
Muge Mutis, Ufuk Beyaztas, and Han Lin Shang build a robust estimator for a spatial regression model where the predictor is a curve. In scalar-on-function regression, a number, such as a pollution level, depends on a whole function, such as a day's profile, plus the pull of neighboring locations. Standard estimators, likelihood-based or mildly robust, can be wrecked by outliers, odd predictor curves, or strong spatial dependence.
Their framework first compresses the functional predictor with a robust version of principal component analysis. It then estimates the spatial model through bias-corrected equations allowing redescending loss functions, which fully reject extreme outliers rather than merely downweighting them. Theory establishes consistency and the limiting distribution of the estimated slope function. Simulations show the method holds up under clean data and beats classical and Huber-type rivals under contamination. An application to French air-quality data improves prediction as well.
Catalog record
| CREATOR | Muge Mutis |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 153 KB |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.00470