ESR-AbstractRater: extended data — prompt templates, detailed statistical tables, STROBE checklist and analysis code (2026)

Extended data for the study Can Local Large Language Models Discriminate Journal Selectivity? A Comparative Evaluation Using Published Radiology Articles (F1000Research, 2026). Files Extended_data_S1_prompts.pdf: verbatim prompt templates used in the abstract-only and full-text conditions. Extended_data_S2_crosstab.pdf: full cross-tabulation of recommendation rates (total, number of yes, percentage) by model, input condition, source journal and target journal. Extended_data_S3_holm_bonferroni.pdf: all 32 McNemar tests with uncorrected and Holm–Bonferroni-corrected p values, by test family. Extended_data_S4_confidence.pdf: verbalised confidence distributions (n, mean, median, mode, SD, proportion at exactly 0.85) by model and condition. STROBE_checklist.pdf and STROBE_checklist.md: completed STROBE checklist (cross-sectional) with the corresponding manuscript sections. code_v1.0.zip: analysis code (Python 3.10) for corpus retrieval, model inference, statistical analyses and figure generation, with its own README describing the pipeline, the software environment and the mapping between scripts and the tables and figures of the article. README.md: description of the files in this record. The raw full texts of the analysed articles are not redistributed, as they remain under publisher copyright; code_v1.0.zip contains the script that re-acquires them from the DOI listed for each article in the underlying data record (10.5281/zenodo.23865251), which also contains the corpus and all 13,008 model responses.