Guides

Use these guides after completing Getting started. Each guide produces one concrete result and links to exact API reference where needed.

Evaluation and thresholds

• Evaluate subgroup performance configures a complete metric audit with support counts and bootstrap intervals.
• Compare models and groups estimates paired score and reference-level contrasts.
• Evaluate calibration and decisions builds reliability summaries and descriptive decision curves.
• Optimize a scalar threshold selects an operating point by Youden index or a sensitivity or specificity target.
• Apply conditional thresholds assigns different operating points to feature levels.

Error analysis and reporting

• Analyze confusion-group representation examines which subgroups are concentrated in TP, TN, FP, and FN sets.
• Export and style results chooses between nested numeric values, display tables, and notebook styling.
• Plot metric confidence intervals creates one Matplotlib interval figure per feature.
• Plot score distributions compares raw score distributions using shared histogram bins.
• Build hierarchical visualizations compiles node arrays and passes them to Plotly.

Extension

• Create a custom metric adds a metric over the stable release’s recognized inputs.