Another Look at Bandwidth-free Inference: a Sample Splitting Approach
Yi Zhang, Xiaofeng Shao · 2026
Read Another Look at Bandwidth-free Inference: a Sample Splitting Approach by Yi Zhang, Xiaofeng Shao (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-2606-00864.txt.
Description
Zhang and Shao fix a flaw in bandwidth-free tests in this 2026 econometrics paper. Such tests judge a multi-part claim about time series data. They are loved for ease: no tuning knob, and size closer to the nominal level than classic methods. But with small or mid samples, size can warp badly. Strong time links and many parameters make the warp worse.
Their cure is sample splitting plus self-normalization. Splitting cuts the parameter down to one dimension for the final test. They build two test forms, one based on the max gap and one on squared gaps. Limit laws are derived under both the null and other cases. Simulations show the size warp shrinks a lot. The trick fits many jobs: mean tests, zero-autocorrelation tests, regression claims, and change-point tests for multivariate means. The pair present it as a broad patch, not a one-case fix.
Catalog record
| CREATOR | Yi Zhang, Xiaofeng Shao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 214 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.00864