Prioritizing incident information for emergency dispatch: Reproducibility archive (code, fitted models and outputs)

Reproducibility archive for the manuscript "Prioritizing incident information for emergency dispatch: A two-dimensional value-of-information framework with a scale-adaptation diagnostic". The study uses London Fire Brigade incident records (London Datastore) to value incident information by when it becomes available and by whether it reduces allocation loss through the forecast level or the forecast uncertainty. A fixed NGBoost location–scale model is evaluated on four nested information scenarios: call time and borough (A), property category (A+), a binary call-nature indicator separating fire-type calls from special services (C), and the final incident group (B). The record contains two archives. ThreeScenario_Archive.zip holds the original three-scenario analysis (A, A+, B): research code, fitted models, 21 verified CSV exports and documentation. FourScenario_Analysis.zip holds the four-scenario analysis: the fitted Scenario C model; losses, coverage and value of information in 2024 H2, 2025 and January–February 2026; controlled location × scale factorial and Shapley decompositions for all scenario pairs and periods (reproducing the archived A-to-B decomposition exactly); the split at the October 2024 LFB automatic-fire-alarm policy change; operational metrics and service-level sensitivity; the test of the scale-adaptation diagnostic; and the scripts that produce the manuscript's tables and figures. To rerun the analyses, extract both archives side by side and supply the combined LFB input cache, which is not deposited; its SHA-256 and required location are given in the README files.