Data and code for: Direct maize–wheat comparison of yield associations with co-located precipitation-deficit–heat concurrence across African cropping systems
Description
This dataset and code archive supports the revised manuscript “Direct maize–wheat comparison of yield associations with co-located precipitation-deficit–heat concurrence across African cropping systems.” The study compares national maize and wheat yield associations with hot–dry concurrence across 11 African countries during 2001–2018. Precipitation deficit, maximum-temperature anomaly, and their continuous product were constructed within the same native NASA POWER cell and active MIRCA crop month before crop-season and national aggregation. An outcome-blind, pair-preserving overlap rule retained 192 of 198 matched country-years (384 crop observations). The locked pooled model includes country × crop effects and linear trends, crop × year effects, and irrigation-composition interactions; uncertainty is evaluated with country-clustered CR1 inference and exhaustive restricted-null wild-cluster inference. The archive contains the frozen analytical specification, implementation addenda, source manifests, SHA-256 checksums, MIRCA-to-POWER mappings, derived native cell-month and crop-season exposures, common-support scores, national yield and irrigation inputs, the locked analytical panel, model and numerical-audit outputs, rainfed-exposure and FAO-56 SPEI-3 measurement sensitivities, an independent numerical validation, and code that regenerates the manuscript tables and figures. The primary result is a conditional national cropping-system contrast and should not be interpreted as an intrinsic physiological crop parameter. The two exposure sensitivities share observations with the primary analysis and are not independent replications. Public source data are attributed to FAOSTAT/Our World in Data, NASA POWER, and MIRCA-OS. Third-party terms remain controlling for source-derived material. The README provides offline primary reproduction, archive verification, figure/table generation, and an optional full source-to-results rebuild.
Files
Steps to reproduce
Download and extract AFM_Reproducibility.zip without changing the folder structure. Install Python 3.12 and create an isolated environment. Run: python -m pip install -r requirements.txt Verify the unchanged deposit: python run_reproduction.py --mode verify Re-run the locked primary model offline: python run_reproduction.py --mode primary Regenerate Tables 1–2, Supplementary Table S6, and Figures 1–2: python run_reproduction.py --mode figures Optional source-to-results reconstruction: python run_reproduction.py --mode full The full reconstruction requires internet access to the NASA POWER monthly regional API and executes concurrence construction, primary estimation, transparency controls, and output generation in that order. Compare the regenerated focal output with r2_primary/results/AFM_R2_locked_primary_decision_v1.json. The expected coefficient is −0.1184851923 and the expected exhaustive wild-cluster p value is 0.037109375. Consult README.md for expected sensitivity results, folder definitions, source-change diagnostics, and interpretation limits.
Institutions
- Kocaeli ÜniversitesiKocaeli, İzmit