Rating–text inconsistency, purchase verification and online review helpfulness

This repository contains supporting data, a fitted text-rating model, computational outputs, tables and figures for a study of rating–text inconsistency, purchase verification and online review helpfulness. The analysis uses 479,493 reviews from six categories in the Amazon Reviews’23 dataset, covering 2019–2022. File descriptions • analysis_derived.csv and analysis_derived.csv.gz: Numerical analysis dataset containing 479,493 review records, including posted ratings, helpful-vote counts, purchase-verification status, text-derived rating scores, discrepancy measures and model covariates. • source_matching_manifest.csv and source_matching_manifest.csv.gz: Source-matching metadata for 620,158 analysis and text-model development records. Fields include product identifiers and SHA-256 hashes of review text. • development_split_manifest.csv and development_split_manifest.csv.gz: Training, validation and test assignments for 140,665 text-model development records, including validation subsets used for calibration and model selection. • text_model.joblib: Fitted text-rating model, including the text vectorizer, classifiers, mixture weight, isotonic calibrator and training configuration. • analysis_outputs ZIP archive: Statistical estimates, covariance matrices, group-specific contrasts, model diagnostics and supplementary computational outputs. • Figures and tables ZIP archive: Manuscript figures and tables. CSV and GZIP formats Each dataset is provided as an uncompressed CSV file and a GZIP-compressed CSV file. These are alternative formats of the same dataset and should not be combined as additional observations. The CSV versions facilitate inspection and import into spreadsheet software. Use the original .csv.gz versions with the supplied analysis scripts. No renaming or recompression is required. Source data and privacy Raw review text and direct user identifiers are excluded from the public files. The manifests retain product identifiers and text hashes to support matching with the original collection. These linkage fields mean that the manifests should not be described as fully anonymous. The original Amazon Reviews’23 collection is available at:https://amazon-reviews-2023.github.io/ Licensing Author-created derived data, metadata, model artefacts, documentation and outputs are provided under CC BY 4.0. Python source code, where included, is provided under the MIT licence. Original third-party content remains subject to the source provider’s terms.