Cited Half-Life Across 26 Knowledge Fields of OpenAlex: Aggregated, Compositional, and Gilyarevskii Method Estimates (2000–2024)

This dataset contains the research data underlying the article "Compositional effect on cited half-life: a comparison of three aggregation methods across the 26 fields of OpenAlex." It provides cited half-life (CdHL) estimates for all 26 knowledge fields and 252 subfields of the OpenAlex taxonomy, computed over 25 base years (2000–2024) from a corpus of 118,464,542 works classified with a topic assignment score ≥ 0.9. Three aggregation methods are compared: (1) the aggregate method, which pools all citations within a field and extracts the median citation age; (2) the compositional method, which computes CdHL per subfield and takes the median of subfield values; and (3) the Gilyarevskii method, which calculates CdHL per source (journal) and aggregates by citation-weighted mean. Temporal stability metrics — coefficient of variation and linear slope — are provided for each field across the full 25-year series. The dataset also includes the proportion of uncited articles per field as a measure of synchronic survival bias, and the full decomposition of the compositional effect for the four fields where the aggregate and compositional methods diverge by more than 1.5 years. Data were extracted from the OpenAlex public snapshot of January 2026 (dataset: publicdb_openalex_2026_01_rm, hosted on Google BigQuery under the Multiobs project). All computations were performed directly on BigQuery using window functions and Sen (1999) linear interpolation for median estimation. Files included: cdhl_by_field_yearbase.csv — CdHL (aggregate method) per field and base year, with Q25, Q75 and IQR. cdhl_by_subfield_yearbase.csv — CdHL per subfield and base year. cdhl_subfields_latest.csv — CdHL per subfield for base year 2024, with citation counts. cdhl_gilyarevskii_by_field.csv — CdHL computed via the Gilyarevskii journal-weighted method, per field, base year 2024. cdhl_with_renorm.csv — CdHL with renormalization by annual publication volume, per field and base year. method_comparison.csv — Side-by-side comparison of the three methods per field, base year 2024, with absolute differences. stability_by_field.csv — Temporal stability metrics (mean, min, max, CV, linear slope, trend classification) for the aggregate CdHL series per field. stability_by_field_composition.csv — Same stability metrics for the compositional CdHL series per field. stability_by_subfield.csv — Stability metrics per subfield. tabela1_cdhl_por_field.csv — Data underlying Table 1 of the article. tabela2_estabilidade.csv — Data underlying Table 2 of the article. tabela3_comparacao_metodos.csv — Data underlying Table 3 of the article. tabela4_decomposicao.csv — Data underlying Table 4 of the article (compositional effect decomposition). tabela4_cdhl_por_subfield.csv — Full subfield CdHL values for the four fields in Table 4. uncited_by_field.csv — Count and proportion of uncited works per field (2000–2024). taxonomy_fields.csv — OpenAlex field taxonomy (26 fields with IDs and names). taxonomy_subfields.csv — OpenAlex subfield taxonomy (252 subfields with IDs, names, and parent field). fields_list.csv — Auxiliary lookup table for field identifiers.