Robust model selection using likelihood as data
Jongwoo Choi, Neil A. Spencer, Jeffrey W. Miller · 2026
Read Robust model selection using likelihood as data by Jongwoo Choi, Neil A. Spencer, Jeffrey W. Miller (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-2602-23355.txt.
Description
Jongwoo Choi and two co-authors propose robust model selection that treats likelihood scores as data. Model choice is central to stats, but standard methods falter when the true data-making process is not among the options. Information rules and Bayesian views, above all, do not quantify doubt about how well each option approximates the true process.
Given K candidate models and n observations, the authors view the n-by-K matrix of negative log-likelihood values as a random data matrix. Each row's expectation equals the vector of Kullback-Leibler gaps between the options and the true process, up to an added constant. A multi-factor normal model then estimates that average and measures its doubt, giving tuned judgments for robust choice under wrong-model cases. The method is easy to compute, clear, and comes with math pledges including steady results.
Catalog record
| CREATOR | Jongwoo Choi, Neil A. Spencer, Jeffrey W. Miller |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 147 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-23355