Dkplucas/Smallruminant-survey-data-pipeline

Data extraction, data cleaning for the crossbreed farmers of small ruminant in Benin from Kobocollect. # Smallruminant-survey-data-pipeline R-based data pipeline for cleaning, recoding, and statistically analyzing KoboCollect survey data on goat crossbreeding practices (Djallonké × Sahelian × exotic breeds) in Bénin. The pipeline covers household identification, reproduction, health/veterinary follow-up, environmental context, and herd composition, then applies Chi-squared and Monte Carlo tests to selected variables. ## Project Structure ``` Data_cleaning/ Sequential cleaning scripts (unite1cleaning.R → unite7cleaning.R) Each script renames raw KoboCollect columns to coded, analysis-ready variables (e.g. 1=Masculin, 2=Feminin) and exports an intermediate file (data1.xlsx → data7.xlsx). Test_CHI2/ Chi-squared independence tests on qualitative variables (education, sex, marital status vs. zone, etc.), built on a shared helper (chi2_shared.R) plus a qualitative-variable extraction script (qualitative.R). Monte_Carlo/ Monte Carlo simulation of Chi-squared tests (chisq.test(..., simulate.p.value = TRUE)) used when expected cell counts are too low for the standard approximation. ``` ## Pipeline Flow 1. **Data_cleaning** — Raw KoboCollect export is processed sequentially through `unite1cleaning.R` … `unite7cleaning.R`. Each script: - Loads the previous stage's output (`dataN.xlsx`) - Renames columns to embed their coding scheme directly in the header (e.g. `Sexe: 1=Masculin, 2=Feminin`) - Fills/recodes blank or free-text responses into standardized numeric codes - Verifies no unexpected empty cells remain - Exports the next stage (`dataN+1.xlsx`) 2. **Test_CHI2** — Consumes the final cleaned dataset to test independence between qualitative variables (e.g. education level, sex, marital status) and geographic zone. 3. **Monte_Carlo** — Re-runs Chi-squared tests with Monte Carlo p-value simulation for variables where expected counts violate the Chi-squared approximation's assumptions. ## Requirements - R (≥ 4.0 recommended) - R packages: `dplyr`, ` …