De-identified analytic data and reproducibility materials for "Beyond views: Video duration and the reach-resource distinction in platform-mediated interaction on Bilibili"

This repository contains the de-identified analytic data and reproducibility materials supporting the manuscript “Beyond views: Video duration and the reach-resource distinction in platform-mediated interaction on Bilibili.” The study analyzes publicly accessible video-level metadata retrieved from Bilibili in June 2025 using four DeepSeek-related searches. The original 1,241 search-result rows were screened and reconstructed into 999 unique-video records for analysis. The repository includes: (1) a de-identified analytic dataset, (2) a spreadsheet containing analysis-ready variables and supplementary robustness results, (3) Python code for reproducing the principal negative binomial models and sensitivity analyses, (4) a README file, and (5) licensing information. Direct identifiers and free-text platform fields, including creator names, video titles, descriptions, and tags, are excluded from the public analytic dataset to reduce re-identification risk. Pseudonymous creator cluster identifiers are retained only to reproduce creator-clustered standard errors. Plays are operationalized as play-based realized exposure rather than unique-user reach. Favorites and shares are analyzed as retention- and circulation-oriented interaction traces and should not be interpreted as direct measures of user motive, engagement experience, learning, or content quality.