Outlier Detection Methodology

Data: {{metric_name}} from {{data_source}}. Distribution: {{distribution}}. Suspected outliers: {{suspected_outlier_pct}}% of data. Concern: {{outlier_concern}}. Analysis context: {{context}}.

Design outlier detection:
1. Visual inspection: histogram/boxplot of {{metric_name}}
2. Method 1 (IQR): identify values outside {{iqr_multiplier}}×IQR
3. Method 2 (Z-score): identify values >{{z_threshold}} std devs
4. Method 3 (Isolation Forest): unsupervised anomaly detection
5. Method 4 (Business logic): {{custom_rule}}
6. Consensus: which values are flagged by ≥2 methods?
7. Action: remove, cap, or segment outliers for {{analysis_context}}?