Reproducibility Package for Point-in-Time Validity in Municipal 311 Service-Time Prediction: A Multi-City Benchmark of Outcome Construction, Information Availability, and Temporal Evaluation

This deposit provides the reproducibility package for the study “Point-in-Time Validity in Municipal 311 Service-Time Prediction: A Multi-City Benchmark of Outcome Construction, Information Availability, and Temporal Evaluation.” The study evaluates how outcome construction, information availability, and temporal evaluation affect municipal 311 time-to-disposition prediction across Austin, Chicago, and New York City. Version 2.0.0 extends the original package with targeted manuscript-revision analyses while retaining the original primary 2 × 2 × 2 factorial benchmark and prespecified sensitivity analyses unchanged. The added analyses include a matched-objective censored-versus-closed-only AFT comparison, a hard terminal/resolution proxy versus mutable non-terminal process-field ablation, a 2,000-replicate paired creation-day block bootstrap for uncertainty-aware model-selection comparisons, and a hard-proxy prevalence audit. The package contains analysis scripts, locked protocol and implementation records, audit outputs, supporting data and compact result tables, revised manuscript figures, environment information, source-portal metadata snapshots, and integrity checksums. Raw municipal request records, request identifiers, request-level predictions, fitted model objects, and feature-table files are not redistributed. The original 256 primary and prespecified sensitivity result cells are retained unchanged. The additional analyses are targeted revision analyses and are not included in that count. The endpoint represents time to an administrative terminal disposition recorded by the source system rather than verified physical resolution. The point-in-time information contract was reconstructed from final public extracts under declared field-stability assumptions rather than archived daily snapshots.