PPDisentangle Data and Results

Companion results and matching source snapshot for PPDisentangle v0.2.0 (release: https://github.com/duncan-clark/PPDisentangle/releases/tag/v0.2.0; source commit: 3eade98b25d104aaa351422dd5c8c24d393aafa3; software DOI: https://doi.org/10.5281/zenodo.23041208). The ZIP contains the exact released software source, retaining its included MIT licence; the results archive is distributed under CC BY 4.0. Updated Oklahoma job 8804859 fits plus ATE backfill: 818 post-treatment events, 250-day temporal truncation, and 512 attempted bootstrap replicates. The paper uses a 100-day observed-AOI-versus-none contrast; saved counts are -51.508 (naive) and 175.200 (SEM). Bootstrap summaries use 507 and 512 retained replicates and translate the full distribution to the fitted contrast. Includes main simulation job 5228509 and the unchanged 67-scenario robustness suite. Figures and tables are regenerated from frozen fits, with source-commit provenance and per-file checksums. No simulations or model fits were rerun for this deposit. The software reproduction guide documents settings and the distinct simulation/application intervention conventions. The results are supplied as one 81.7 MB tar.gz archive. All 73 simulation RDS files were checked during lossless recompression: their decompressed R serialization streams are byte-for-byte identical to the original larger archive. No saved fields, replicates, diagnostics or plot objects were removed. The standard readRDS() and v0.2.0 reproduction commands are unchanged. compression_manifest.json records the before/after sizes and matching serialization hashes.