statistical-modeling

Survival analysis for biomedical biomarker discovery is performed using Cox proportional hazards regression and scikit-survival models. The skill enforces rigorous cross-validation protocols and confidence interval reporting to ensure reproducible scientific output.

Interpreting machine learning models for biological meaning is best achieved using SHAP values and permutation importance. This ensures model decisions remain transparent and biologically meaningful during classification and regression tasks.

Mixed-effects modeling with imbalanced biomedical datasets is supported through statsmodels and pingouin. The skill applies appropriate metric selection and standardized evaluation protocols to handle data imbalance effectively.

Avoiding data leakage and overfitting in classification tasks requires strict cross-validation protocols and standardized evaluation metrics. The skill enforces these best practices to maintain statistically sound and auditable biomedical analysis.

Mixed-effects models are needed instead of standard regression when your research data involves hierarchical structures or repeated measures. The skill provides a structured decision tree for method selection based on data type and research goals.

The statistical modeling workflow supports feature selection and model evaluation together within complex biomedical research pipelines. It integrates method selection, cross-validation, and interpretability reporting for comprehensive reproducible analysis.