Sharp transport identification with unknown Gaussian covariance
Scott Schubr, Luminary Oddities — October 27, 2026

Contents
main.pdf: complete manuscript (15 pages).
main.tex: complete editable LaTeX source, including bibliography.
imsart.cls, imsart.sty: unmodified official IMS article support files.
LICENSE: LaTeX Project Public License for those support files.
CITATION.cff, citation.bib, citation.ris, citation.json: article citation records.
build.py: portable local LaTeX build wrapper.
MANIFEST.sha256: SHA-256 checksums of every other public member.

Build
Compile main.tex with pdfLaTeX or LuaLaTeX, or an existing Tectonic executable.
The source uses the current official ejsv2 and documented noshowframe options.
Keep imsart.cls and imsart.sty beside main.tex. Standard packages loaded by
the source and IMS support files are required. The bibliography is included
directly; BibTeX is unnecessary. No data or numerical producer is required.
Run pdflatex main.tex twice, or python build.py --engine tectonic, replacing
the engine argument with an executable path if needed.
Official support: https://github.com/vtex-soft/texsupport.ims_cosponsored-ejs
The unmodified support files are redistributed under the included LICENSE.

The paper proves sharp transport partial identification under approximate
latent covariance information, a minimax primary/calibration sample tradeoff,
full smooth-class estimation from paired independent-error observations,
exact compatible laws for Gaussian-divisible references, and rough-input
fixed-reference testing. Its known-noise analytic inverse theorem is
explicitly stated and cited. The complete covariance-transfer and statistical
arguments appear in the article, together with its hypotheses and constants.

Checksums
Use a SHA-256 tool to compare files with MANIFEST.sha256.
The manifest excludes itself.
