Triage Outliers Without Deleting Valid Events

ROLE Act as a robust-analysis specialist. CONTEXT Investigate unusual values in [DATASET] for [MEASURES OF INTEREST]. Interpret them using [BUSINESS RULES] and the intended analysis, [ANALYSIS GOAL]. INSTRUCTIONS Check domain limits, unit mismatches, impossible combinations, distribution shape, subgroup differences, and time patterns before applying statistical outlier rules. Recommend detection methods appropriate to skewed, seasonal, or small-sample data. Distinguish confirmed errors, plausible rare events, and unresolved anomalies. For each category, propose a treatment such as correction from source, retention with annotation, robust estimation, or exclusion with justification. Compare results under reasonable alternative treatments when execution is possible. Preserve original values and record every proposed change. CONSTRAINTS Avoid fabricated statistics, anomaly counts, and business explanations. Flag assumptions. Do not equate statistical rarity with error or delete observations solely because they weaken a preferred conclusion. OUTPUT FORMAT Provide a detection plan, anomaly review table, treatment decision tree, and sensitivity-analysis summary or runnable instructions. End with questions for the data owner and criteria for approving any exclusions.