English | Español

---

Technical reference for run_analysis.py: phases, inputs, outputs, and artefacts.

---

• Script: scenarios/analysis/run_analysis.py
• Run from: Repository root (parent of scenarios/).
• Example: python3 scenarios/analysis/run_analysis.py --corpus corpus_v1 --phase all

---

Phase	Purpose	Main inputs	Main outputs
features	Extract feature vector from each .settings (world_area, aspect_ratio, …)	corpus dir (e.g. corpus_v1)	data/features.csv, data/scenario_list.txt
features_report	List used features + settings not used (with reasons)	corpus, data/	reports/features_report.txt, reports/features_report.md
normalize	Z-score per column (ignoring NaN), then impute NaN→0; write core/reduced subsets	data/features.csv	data/features_normalized.csv, data/normalization_params.csv, data/features_core.csv (23), data/features_reduced.csv (17)
correlation	Pearson, Spearman, cosine & Euclidean distance between scenarios	data/features_normalized.csv	data/correlation_*.csv, data/distance_*.csv, reports/correlation_report.txt, reports/multiple_comparisons_report.txt, reports/clustering_report.txt, data/cluster_assignments.csv
feature_correlation	Feature–feature correlation matrix (core 23×23)	data/features_core.csv	data/feature_feature_correlation_core.csv, figures/heatmap_feature_feature_core.*, reports/feature_feature_correlation_report.txt
ablation	Compare diversity metrics for 17 vs 23 vs 46 features	data/features_normalized.csv, features_core.csv, features_reduced.csv	reports/ablation_report.txt, data/ablation_metrics.csv
figures	Heatmaps, histograms, PCA scatter	data/*.csv (from correlation)	figures/*.png, figures/*.pdf
output_metrics	Build output_metrics from ONE reports	*_MessageStatsReport.txt (e.g. in reports/)	data/output_metrics.csv
outputs	Correlation/distances on output vectors	data/output_metrics.csv	data/*_outputs.csv, reports/outputs_correlation_report.txt, figures/heatmap_pearson_outputs.*

---

• --corpus <path> — Path to corpus directory (required for features; can be relative, e.g. corpus_v1 if run from repo root with scenarios/corpus_v1).
• --phase <name> — One of: features, features_report, normalize, correlation, feature_correlation, ablation, figures, output_metrics, outputs, all. With all, runs features → features_report → normalize → correlation → feature_correlation → ablation → figures → output_metrics (not outputs).
• --reports-dir <path> — For output_metrics: directory containing ONE report files (default: reports/ at ONE root).
• --threshold 0.7 — Correlation threshold for reports (default 0.7).
• --strict — Require 100% of pairs with |r| < threshold (for correlation phase).

---

Location	Contents
analysis/data/	features.csv, features_normalized.csv, normalization_params.csv, features_core.csv (23), features_reduced.csv (17), correlation_pearson.csv, correlation_spearman.csv, correlation_pearson_pvalues.csv, distance_cosine.csv, distance_euclidean.csv, feature_feature_correlation_core.csv, ablation_metrics.csv, cluster_assignments.csv, output_metrics.csv, *_outputs.csv
analysis/figures/	heatmap_pearson., heatmap_spearman., heatmap_feature_feature_core., histogram_correlations_., scatter_pca_regression., scatter_max_r_pair_regression., heatmap_pearson_outputs.
analysis/reports/	correlation_report.txt, multiple_comparisons_report.txt, feature_feature_correlation_report.txt, ablation_report.txt, clustering_report.txt, scenarios_to_diversify.txt, features_report.txt, features_report.md, outputs_correlation_report.txt, observaciones_correlacion.md, plan_radical_scenarios.md

---

• Quickstart — How to run the pipeline
• Methodology — How features and correlation are defined
• Features reference — All 46 features (description, origin); core 23 vs extended; settings not used (with reasons)
• Results overview — Main results