Research record for: Mid-Infrared Spectroscopy and Machine Learning for Plastic Classification: A Critical Review (search record, coding, statistics, figure scripts and reliability materials)

Research record for the critical review "Mid-Infrared Spectroscopy and Machine Learning for Plastic Classification: A Critical Review of Preprocessing-Model Coupling, Validation Realism, and Deployment Readiness" (Abdelghani, Bar-Ziv and Long; Critical Reviews in Analytical Chemistry, manuscript BATC-2026-0510). The archive holds the primary material behind the numbers in the review: the database exports of 4 and 5 April 2026 (Web of Science, Scopus, IEEE Xplore) and the screening decision for every record, with the script that derives the PRISMA 2020 counts; the 58-variable extraction workbook for the 103 coded studies and one evidence sheet per study with the quotation and locator behind every coded value; the coding rules and the revision rulings with the cells they changed; the statistics file that every table and figure reads from, by evidence tier and affiliation set, with Wilson intervals, and the scripts that build it; the figure scripts; and the reliability materials (rater instrument and packet text, sample selection, the May 2026 pilot and its recomputation, the blind re-read of 20 coded cells and the full-read screen of the CORE texts). See README.md inside the archive for the folder map, the reproduction order and the few inputs that cannot be deposited. Version 2 (28 October 2026) corrects version 1. Version 1 contained a folder, 05_reliability/independent_rating_returns, presented as the return of the independent rater; its scoring sheet was generated by an AI model on 28 October 2026 and is not an independent rating by a person, so the folder is withdrawn and nothing in it is used. The independent rating will be done by a person and added to a later version. Version 1 also contained the contact author's telephone number and 738 e-mail addresses of other people (in the Scopus exports of 27 October 2026 and in quotations from the papers); they are removed from this version. Six scripts and records carry corrections made on 28 October 2026. README.md in the archive lists every change. Version 3 (29 October 2026) adds the independent rating of the reliability sample (05_reliability/independent_rating_returns): the scoring sheet, scoring form and questions log as returned by the rater on 29 October 2026, the agreement statistics computed against the reference coding frozen on 28 October 2026, and the rater's written account with names and addresses removed. A completed scoring sheet produced with a generative AI tool had been sent to the rater as an example before the return; the rater states that it was seen only after the scoring and was not used. Ruling V-35 applies the rule, fixed before the rating, that the rater's value governs the disputed cells of the author-affiliated papers. README.md in the archive lists every change. Version 4 (31 October 2026) reduces the database exports of Web of Science, Scopus and IEEE Xplore to the record identifiers and the basic bibliographic fields, removing the abstracts, keywords, addresses and the other fields, and removes the abstracts from two files of the verification re-run; the scripts give the same counts on the reduced files. It adds a rights note: the CC BY 4.0 license covers the authors' own work only, not the database records, the quotations from the reviewed papers or the independent rater's returned materials. It adds the inputs that the deposited scripts read and that version 3 lacked: the paper index, the screening flags of April 2026, the map of bibliography keys, the script and the reference statistics of the agreement analysis of the independent rating, and the statistics of the October 2026 instrument pilot. It corrects the DOIs of two records and the years of four in the search record without changing any count, reports the co-author's check of the pre-screen sample and the use of generative AI in coding the four recovered candidates, follows ruling V-35 in the rule file of Table 2, states the affiliation rule of the review in the statistics file, removes the name of a student organization from the rater's written account, and describes more narrowly what the record lets a reader reproduce. README.md in the archive lists every change.