Brief — Part IX "Empirical Results" (file manuscript/sections/part09_results.tex; bib manuscript/references_part09.bib)
Target 40–60 pages: ALL verified quantitative findings, with figures and tables, organised by research question RQ1–RQ10 and hypothesis H1–H8 (protocol/PROTOCOL.md P1). Follow WRITER_COMMON.md. Sources (corpus v2, classification v1.0):
- Corpus & taxonomy: data/metadata/prisma_counts.json, linkage_report.json, linkage_sample_summary.json (H-1 auto precision), data/processed/classification/classification_report.json (copy in reports/), taxonomy/validation/metrics.json (+ disagreements.csv counts).
- Volume/composition (RQ1): results/bibliometrics/volume_yearly.csv, volume_monthly.csv, results/graph/category_counts_yearly.csv, category_counts_monthly.csv, results/bibliometrics/cagr_by_category.csv, diversity_yearly.csv, category_share_ci_yearly.csv.
- Structural change (RQ2, H1, H3): results/econometrics/its_results.json (series total and cat_), breaks.csv, placebo_*.csv, reports/PHASE4_STRUCTURAL_CHANGE_TOTAL.md (preliminary reading).
- Influence (RQ3): results/bibliometrics/citation_windows_yearly.csv, top_cited_by_year.csv, literature_synthesis/landmarks/selection_report.json.
- Networks (RQ4–RQ6, H4–H6): results/graph/{field_flows_yearly, category_flows_yearly, area_flows_yearly, cross_area_share_yearly, modularity_yearly, category_cooccurrence_yearly, cited_age_yearly, reference_coverage_yearly}.csv; results/bibliometrics/{team_size, international_collab, industry_share, countries, venue_hhi}_yearly.csv (Part VIII covers the actor networks in depth: summarise and cross-reference).
- Semantic (RQ5, H4 semantic level): results/semantic/{category_drift.csv, category_similarity_baseline.csv, category_similarity_post.csv, clusters.csv, stability.json} if present (else \pending{semantic atlas}).
- Openness/reproducibility (RQ9, H7): volume_yearly.csv columns n_arxiv, n_oa, n_preprint, n_abstract; top_sources_by_period.csv.
- Emerging detection (RQ10, H8): results/emerging/metrics_h12.json, metrics_h24.json, predictions.csv, known_set_lead_times.csv.
- Figures in manuscript/figures/ (see figures/FIGURES.json for captions/sources) and tables in manuscript/tables/ (TABLES.json):
include every relevant one with \includegraphics / \input and a \label; you may add tables to tables/make_tables.py (append only) and re-run it.
Rules: every number through macros (
python3 tables/make_numbers.py .first; macro names derive from JSON keys and CSV columns — e.g. \PrismaCsCoreEligible, \ItsSeriesTotalItsChatgptSlopeChange is NOT generated; for ITS numbers use the generated\ItsSeries...macros if present, else add the table and cite it) or generated tables. State each hypothesis, what the test shows, and the verdict (supported / not supported / inconclusive) with uncertainty; null and opposite results are reported as such; no causal language; right-truncation and date-precision caveats on every time-series statement. Flag with \pending{H-4} the interpretive statements that await PI review.