Species-Specific Economic Burden of Foot-and-Mouth Disease in Mixed-Species Smallholder Livestock Systems: Evidence from Samsun, Turkiye
This repository contains the complete reproducibility package accompanying a manuscript that quantifies and statistically compares farm-level, FMD-attributable economic losses across cattle, water buffalo, sheep, and goats on 286 mixed-species smallholder farms in Samsun Province, Turkiye, surveyed during the 2022-2023 production period -- a period that coincided with the officially documented 2023 incursion of FMD serotype SAT-2.
This package is intentionally organized around the study rather than any single journal submission. If the manuscript title, framing, or target journal changes during peer review, this repository and its contents remain valid without modification.
This repository follows open science and computational reproducibility principles and includes:
- Complete Python source code (descriptive statistics, non-parametric tests, farm-clustered regression, spatial autocorrelation analysis, figure generation)
- The complete farm-level analytical dataset (N = 286 mixed-species smallholder farms, 2022-2023)
- Village-level geographic coordinates used for the spatial analysis
- Comprehensive documentation of data provenance, sampling design, and known data-quality handling
- Software environment specifications
FMD-Multispecies-Economics-Samsun/
├── code/
│ ├── _paths.py # shared path/config
│ ├── 01_sample_overview.py # Table 1
│ ├── 02_demographics.py # Table 2
│ ├── 03_species_disease_comparison.py # Table 3; Figure 2; KW/MW tests
│ ├── 04_regression_analysis.py # Table 4; Figures 3-5; VIF; robustness
│ ├── 05_income_vulnerability.py # Figure 6; Section 3.5 statistics
│ ├── 06_mortality_analysis.py # Table 5; Figure 7; consistency check
│ ├── 07_spatial_analysis.py # Table 6; Figures 8-9 (Moran's I)
│ ├── 08_benchmark_table.py # Table 7
│ └── run_all.py
├── data/
│ ├── raw/
│ │ └── village_coordinates.csv
│ └── processed/
│ └── farm_level_data.csv # N=286 farms, main analytical dataset
├── output/ # generated tables (.csv)
├── figures/ # Figure 1 (map) + generated Figures 2-9 (.png, 300 DPI)
├── docs/
│ ├── CODEBOOK.md
│ ├── DATA_DESCRIPTION.md
│ ├── REPRODUCIBILITY_CHECKLIST.md
│ └── Replication_Guide.md
├── README.md
├── CHANGELOG.md
├── CITATION.cff
├── .zenodo.json
├── LICENSE
├── requirements.txt
├── environment.yml
└── .gitignore
- docs/CODEBOOK.md -- analytical workflow and script-by-script description
- docs/DATA_DESCRIPTION.md -- data provenance, sampling design, full variable dictionary, and known data-quality notes
- docs/REPRODUCIBILITY_CHECKLIST.md -- reproducibility checklist
- docs/Replication_Guide.md -- complete, step-by-step replication guide
Before reusing this dataset, read docs/DATA_DESCRIPTION.md in full.
This is original human-subjects survey data, not third-party public
data; the document covers data governance, the distinction between two
similarly-named income fields, the sampling-stratum field, and several
data-quality issues identified and corrected during analysis.
conda env create -f environment.yml
conda activate fmd-multispecies-econ-reproor
pip install -r requirements.txtcd code
python run_all.pyThis reproduces the complete analytical workflow: the sampling-stratum overview (Table 1), household socio-demographic characteristics (Table 2), the species-level FMD occurrence and loss comparison with Kruskal-Wallis and Bonferroni-corrected pairwise Mann-Whitney tests (Table 3, Figure 2), the farm-clustered regression with VIF diagnostics and trimmed-sample robustness check (Table 4, Figures 3-5), farm-level income vulnerability and the off-farm-income-diversification analysis (Figure 6), cattle age-sex-class mortality with its data-consistency check (Table 5, Figure 7), the Moran's I spatial autocorrelation analysis across five neighborhood sizes (Table 6, Figures 8-9), and the benchmarking comparison against published estimates (Table 7).
Expected runtime: well under one minute on a standard laptop. The
slowest step is the 999-permutation Moran's I inference in
07_spatial_analysis.py.
Figure 1 (the Samsun Province study-area map) was constructed as a
cartographic base map rather than a Python script output; see
docs/CODEBOOK.md for details. The finished figure is provided directly
in figures/Figure1_study_area.png.
| Script | Produces |
|---|---|
01_sample_overview.py |
Table 1 (stratum distribution) |
02_demographics.py |
Table 2 (socio-demographic characteristics) |
03_species_disease_comparison.py |
Table 3 (FMD occurrence and losses by species); Figure 2; Kruskal-Wallis and pairwise Mann-Whitney tests |
04_regression_analysis.py |
Table 4 (regression); Figures 3 (species effects), 4 (herd/farm covariates), 5 (prevalence effect); VIF diagnostics; trimmed-sample robustness check |
05_income_vulnerability.py |
Figure 6 (income-vulnerability distribution); farm-level income statistics; buffalo-holding and off-farm-income comparisons |
06_mortality_analysis.py |
Table 5 (mortality by age-sex class); Figure 7; mortality data-consistency check; chi-square and two-proportion z-tests |
07_spatial_analysis.py |
Table 6 (Moran's I by k); Figure 8 (prevalence map); Figure 9 (Moran scatterplot) |
08_benchmark_table.py |
Table 7 (benchmarking against published estimates) |
| (static base map; see docs/CODEBOOK.md) | Figure 1 (Samsun Province map) |
The analytical dataset intentionally carries two distinct off-farm
income measures; using the wrong one will not reproduce the reported
statistics. This is documented in full in docs/DATA_DESCRIPTION.md:
nonfarm_income("Tarim disi gelir") -- income from sources entirely outside agriculture. This is the field used to define "off-farm income source" throughout the manuscript (Table 2: 33.9% of farms; Section 3.5 diversification analysis).offfarm_agri_income("Isletme disi tarimsal gelir") -- agricultural income earned off the respondent's own farm (e.g., day labor on other farms). Not used in the manuscript's off-farm-income statistics; including it changes the reported figure from 33.9% to 37.4%.
Every script in this package was run and its output compared table-by-
table against the manuscript before this package was finalized. Tables
1, 4, 5, and 6 reproduce exactly. Table 2, Table 3, and the income-
vulnerability statistics in Section 3.5 reproduce to within
approximately 1% (minor rounding/ordering differences from the original
iterative analysis; no reported significance level, effect direction, or
ranking is affected). Full detail in docs/CODEBOOK.md and
docs/DATA_DESCRIPTION.md.
Please cite both the published article (once available) and this
archived repository. Citation metadata are provided in CITATION.cff
and .zenodo.json.
MIT License (code in this repository). The underlying farm-level survey
dataset is original human-subjects research data governed separately;
see docs/DATA_DESCRIPTION.md for provenance, anonymization, and terms
of reuse, and docs/REPRODUCIBILITY_CHECKLIST.md for the ethics-approval
confirmation required before public deposition.
Halil Tosun
Department of Animal Science, School of Agricultural and Food Sciences, ADA University, Baku, Azerbaijan
ORCID: https://orcid.org/0000-0001-5117-0390
Email: halilibrahimtosun@gmail.com
Hatice Turkten
Department of Agricultural Economics, Faculty of Agriculture, Ondokuz Mayis University, Samsun, Turkiye
ORCID: https://orcid.org/0000-0003-2037-7756
Zenodo DOI: https://doi.org/10.5281/zenodo.21650221
The manuscript's own DOI (once published) will be added to this file and to the citation metadata files at that time.
Version: 1.0.0