Separate fits and joint compatibility in heterogeneous coverage: finite-label models with shared calibration

A mix of human-written and AI-generated textHuman understanding: some partsmath.ST — Statistics Theorymath.PR — Probability

Contributed by HUANG Man ↗

Version 1 / Nov 21, 2026 / CC BY 4.0

Abstract

Two preparation experiments can each fit a heterogeneous coverage model and still fail a shared-environment explanation. We characterize joint compatibility for finite-label probability representations with half-step source probes and pointwise calibration error. Absolute weaker-source void probabilities require the weighted calibration marginals to share mass at least κ=(β1+β2−1)+\kappa=(\beta_1+\beta_2-1)_+. Combining this observed budget with exposure moments and a shared calibration band gives an exact criterion for the closure of the representation. A deterministic-anchor example requires error radius 0.103760.10376 for separate fits and 0.330480.33048 for joint compatibility. On an explicit three-moment family, matching bounds identify the first-order heterogeneous coefficient κ\sqrt\kappa; four environments preserve the anchors and give second-order stability of the minimum radius over unknown centers. A calibration-preserving moment repair and response-opening geometry yield sharp local inverse bounds with unknown, unequal group masses and first anchors. For prescribed mass and standard-deviation envelopes pn,hnp_n,h_n with hn2≤bpnh_n^2\le bp_n, the matching radius-estimation risk is min⁡{1,[n−1/2+min⁡(hn,n−1/4)]/pn}\min\{1,[n^{-1/2}+\min(h_n,n^{-1/4})]/p_n\}. Uniform consistency has a corresponding two-scale necessary and sufficient criterion. Unknown envelopes introduce distinct adaptation costs. An interval honest over the full envelope union has matching expected diameter min⁡{1,n−1/4/pn}\min\{1,n^{-1/4}/p_n\} even on the deterministic subclass. Under uniform protection of the deterministic regular risk, point estimators pay a sharp logarithmic cost elsewhere: one scale-free thresholded projection attains min⁡{1,[n−1/2+min⁡(hn,(log⁡(en)/n)1/4)]/pn}\min\{1,[n^{-1/2}+\min(h_n,(\log(en)/n)^{1/4})]/p_n\}, and a constrained-risk lower bound matches it. The larger mass-unrestricted moment class admits no uniformly consistent radius estimator. Simultaneous Bernoulli confidence events also give finite-sample compatibility certificates.

Author comments: Published English–Chinese preprint: https://doi.org/10.5281/zenodo.24412395. Joint compatibility, calibration-radius stability and inference for heterogeneous finite-label coverage with shared calibration. Companion preprints: doi:10.5281/zenodo.23912425 and doi:10.5281/zenodo.23911961. Reproducibility scripts and synthetic fixtures are included under anc/.

Provenance statement

This paper studies joint compatibility of two preparation experiments in a finite-label heterogeneous coverage representation with shared calibration. It includes analytic proofs, explicit finite-environment constructions and deterministic numerical diagnostics. OpenAI GPT-6.1 Sol and xAI Grok 4.7 assisted with literature synthesis, theoretical and proof development, drafting, mathematical and code checks, and language revision. Additional AI assistance prepared the repository source package. HUANG Man is the sole contributor. Verification included AI-assisted proof review and deterministic replays of finite diagnostics. The bibliography has 24 entries. The historical summary of Heckman, Robb and Walker (1990), reference [10], draws on the publisher abstract. The detailed comparison of moment conditions and test statistics draws on Sections 2–5 of Robb, Frydman and Robertson (2017), reference [11]. The companion preprints are cited by their public version-1 Zenodo DOIs 10.5281/zenodo.23912425 and 10.5281/zenodo.23911961. The source archive includes reproducibility code and synthetic fixtures under anc/. The scripts check explicit finite constructions and certificates. The general projection estimators are specified as variational programs in the manuscript. The English–Chinese edition appeared on Zenodo on 21 November 2026 (version DOI https://doi.org/10.5281/zenodo.24412395).

Tools used

OpenAI
ChatGPTVersion GPT-6.1 Sol
xAI
GrokVersion 4.7

References

No bibliography entries.

Version history

  1. v1 · 53 pagesSubmitted by Man HuangPublished English–Chinese preprint: https://doi.org/10.5281/zenodo.24412395. Joint compatibility, calibration-radius stability and inference for heterogeneous finite-label coverage with shared calibration. Companion preprints: doi:10.5281/zenodo.23912425 and doi:10.5281/zenodo.23911961. Reproducibility scripts and synthetic fixtures are included under anc/.CurrentNov 21, 2026