SHIFT : Robust Double Machine Learning for Average Dose-Response Functions under Heavy-Tailed Contamination
Eichi Uehara · 2026
Read SHIFT : Robust Double Machine Learning for Average Dose-Response Functions under Heavy-Tailed Contamination by Eichi Uehara (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-2605-00176.txt.
Introduction
Eichi Uehara builds a statistical estimator for dose-response curves that can survive badly contaminated data. Standard double machine learning relies on local smoothers, and a single outlier inside a smoothing window can bend the whole curve there. Heavy tails make that weakness serious in practice.
His estimator, SHIFT, combines cross-fitted orthogonalization with a robust second stage and a careful refit on inliers only. The cutoff for inliers is scaled by a robust spread measure taken after the robust fit, a choice that sharply cut error in a focused contamination test. Across many simulation fits, SHIFT recovered the set of true outliers almost exactly while staying competitive on curve shape. A diagnostic suite helps users judge which robust method their data calls for.
Catalog record
| CREATOR | Eichi Uehara |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 190 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.00176