Crop Input Allocation Across Maize and Cocoa: Analysis Data and Reproducible Optimization Code

Published: 2 November 2026| Version 1 | DOI: 10.17632/46hvf9hzc5.1
Contributor:

Description

This archive contains the Python analysis code and derived data supporting a constrained crop input allocation study. It includes held-out maize treatment means and predictions, selected policies, bootstrap scenarios, capacity and price sensitivity results, and paired confidence intervals based on site and experimental-block resampling. The cocoa files contain paired site summaries and descriptive protocol-allocation outcomes. The original maize and cocoa research datasets are available separately at DOI 10.5061/dryad.66t1g1k2g and DOI 10.5281/zenodo.17484584; their raw files are not redistributed here. The nitrogen budgets and improved-protocol slot limits are analytical decision settings. Cocoa protocol comparisons are observational, and healthy pod mass is not interpreted as dry-bean yield or economic return. The README identifies the required source files, directory structure, software dependencies, and execution order.

Files

Steps to reproduce

Extract the archive. Download the four maize workbooks named in README.md from DOI 10.5061/dryad.66t1g1k2g and place them in data/maize/. Download 3_data_percountry.zip from DOI 10.5281/zenodo.17484584 and place the unmodified file at data/3_data_percountry.zip. From the archive root, use Python 3.12 and run: python -m pip install -r requirements.txt. Run the scripts in this order: analysis_maize.py, analysis_capacity_extension.py, analysis_strengthening.py, analysis_cocoa.py, analysis_inference.py, and analysis_incremental_inference.py. For the split-application cost sensitivity, create results/split15/ and rerun analysis_maize.py with CROP_SPLIT_COST=15 and CROP_OUT=results/split15. Compare the generated files in results/ with the supplied outputs. SHA256SUMS.txt verifies the files distributed in this archive.

Institutions

Categories

Engineering, Decision Science, Agriculture

Licence