Supplementary mateial for "Farm-Grouped Multi-Model Analysis of Dairy Herd Improvement Variables Associated with Elevated Somatic Cell Count in Holstein Cows Using Interpretable Machine Learning"

This record provides the supplementary materials and machine-readable results accompanying the revised manuscript: "Farm-Level Grouped Multi-Model Analysis of Dairy Herd Improvement Variables Associated with Elevated Somatic Cell Count in Holstein Cows Using Interpretable Machine Learning." The deposit includes Supplementary Tables S1–S8 in formatted DOCX and machine-readable CSV formats, Supplementary Figures S1–S4 in PDF format, and the Python script used for external corroboration with an independent Belgian dataset. Contents Supplementary Materials.docx: formatted Supplementary Tables S1–S8.TableS1_farm_descriptives.csv: Table S1, farm-level descriptive statistics. Farm identifiers are anonymized consistently as Farm 1 through Farm 37.TableS1_farm_descriptives_overall.csv: Table S1, overall summary corresponding to the farm-level statistics.TableS2a_vif_16var.csv: Table S2, variance inflation factors for the full 16-variable model.TableS2b_vif_14var.csv: Table S2, variance inflation factors for the 14-variable primary model.TableS3_hyperparameter_sensitivity.csv: Table S3, hyperparameter sensitivity results (12 XGBoost and 12 random forest configurations).TableS4_scc_threshold_sensitivity.csv: Table S4, SCC threshold sensitivity results.TableS5_lofo_cv.csv: Table S5, leave-one-farm-out cross-validation results.TableS6_dim_stage_shap.csv: Table S6, SHAP results across lactation stages (14-variable primary model).TableS7_fat_fpratio_sensitivity.csv: Table S7, sensitivity to removal of fat and fat-to-protein ratio predictors.TableS8_imputation_sensitivity.csv: Table S8, imputation sensitivity results.FigureS1_calibration.pdf: Supplementary Figure S1.FigureS2_farm_shap_heatmap.pdf: Supplementary Figure S2.FigureS3.pdf: Supplementary Figure S3.FigureS4_monthly_scc.pdf: Supplementary Figure S4.validate_residual.py: analysis script for external corroboration using the Belgian dataset. Data availability The raw Korean dairy herd improvement test-day records are not included because they are the property of the Korean Dairy Cattle Improvement Center and were provided under a restricted research data-use agreement. The independent Belgian dataset used for external corroboration is not duplicated in this record. It is publicly available under the Creative Commons Attribution 4.0 International license at https://doi.org/10.5281/zenodo.8263430. To reproduce the external corroboration analysis, place the Belgian data_table.csv file in the same directory as validate_residual.py before running the script. Version note This version provides the final revised supplementary materials, aligns all machine-readable CSV file names with the manuscript table numbers (Supplementary Tables S1–S8), and replaces farm names with sequential anonymized identifiers (Farm 1–Farm 37).