PERG slope figure-generation and analysis code

Abstract

Reproducible analysis and figure-generation code supporting the manuscript "PERG slope features for distinguishing normal and abnormal clinical records in a PhysioNet secondary analysis" by Ying-Jen Chen and Yi-Lin Chiu. This archive contains the code and frozen numerical inputs supporting Figures 1-6, supplementary Figures S1-S5, and Tables 1-3 and S1-S3. It includes the model-feature comparison, SHAP interpretation, record-wise calibration and uncertainty analyses, linked-visit grouped validation, and normal-reference sensitivity analysis. A figure/table-to-code map, dependency records, reproduction instructions, and file checksums are supplied. The analytical code and result files are unchanged from the current Additional File 2; only archive documentation, citation metadata, and licensing have been supplemented. Earlier figure layouts, manuscript-editing tools, correspondence, and process logs are not included. The source is the publicly available PERG-IOBA Dataset version 1.0.0 on PhysioNet (https://doi.org/10.13026/d24m-w054). Raw waveforms must be downloaded separately and are not redistributed. The archive includes derived record-level features, de-identified record links, and numerical summaries. These derived data retain the source Open Data Commons Attribution License and attribution requirements; see DATA_LICENSE.txt. The MIT License applies to author-maintained program code, not to the source dataset or its derived data. Record-wise model selection, fixed-model grouped comparisons, model explanations, and calibrated operating metrics are distinct analyses. The grouped comparisons do not perform nested model reselection, and the archive does not establish external validation or clinical disease-exclusion performance. See README.txt for replay limitations. Funding: Ying-Jen Chen was supported by Tri-Service General Hospital (TSGH-D-113128 and TSGH-D-115131), the Ministry of National Defense (MND-MAB-D-113073), and Taoyuan Armed Forces General Hospital (TYAFGH_E_114049). Yi-Lin Chiu was supported by the National Science and Technology Council (NSTC-115-2628-B-016-006-MY3 and NSTC 114-2314-B-016-024) and the Ministry of National Defense (MND-MAB-D-115176). Software archive DOI: https://doi.org/10.5281/zenodo.20683325. Version 1.1.0.