Archived experimental results for: SARIMAX.jl: Open-Source Optimization-Based SARIMAX Modeling in Julia

Archived experimental results of the article SARIMAX.jl: Open-Source Optimization-Based SARIMAX Modeling in Julia, submitted to the International Journal of Forecasting (special issue Advances in Open Source Forecasting Software): for every experiment behind the values the article prints, the raw results and run logs, the files the table generators read, the tables as they were generated at the time, the original runners and environment files, and a provenance record of every file. It is the input of ./reproduce.sh tables in the Sarimax.jl repository (branch ijf-2026-replication, replication/ijf-2026), which rebuilds the tables from these files and compares each cell with its historical value. M4 competition data and PJM load data are not redistributed (their licences are not established); the replication package downloads them from pinned sources and verifies them. See README.md inside the bundle and LIMITATIONS.md in the repository. The record also contains the replication container image (linux/amd64), exported with docker save from ghcr.io/lampspuc/sarimax-ijf-replication@sha256:ffb9d6bbde694cd9227ff761d232dc6b731c45abc924b2e7bdf3961b746f1637; load it with docker load -i sarimax-ijf-replication-v1.0.0-linux-amd64.tar.gz. The CC BY 4.0 licence of this record applies to the experimental results; the software inside the image keeps its own licences, listed in /opt/licenses of the image. The image does not include SCIP 8.0.0 (ZIB Academic License), which ./reproduce.sh fetch scip installs.