Outlier and Anomaly Investigation Guide
Separate errors from real extremes.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Separate errors from real extremes in my data. Dataset and purpose: [DATASET_AND_PURPOSE]. The suspicious values and context: [SUSPICIOUS_VALUES]. Tool: [TOOL]. For each outlier, follow this tree and record the decision: 1. Is it possible? Check physical or business limits (a negative quantity, an age of 250, a date in the future). IF impossible THEN treat as an error; trace its source. 2. Is it a known data issue (unit mix-up such as dollars versus cents, a placeholder like 9999, a duplicated row, a decimal slip)? IF yes THEN correct with a documented rule. 3. Can it be verified against another source (invoice, system record, person)? IF yes THEN verify. 4. Is it plausible but extreme? IF verified THEN keep, and report results with and without it. 5. Is it influential on the conclusion? Show how the average, total, or trend changes with and without it. 6. Document the decision (kept, corrected, excluded), the reason, who decided, and the date. Provide formulas or code for my tool to detect candidates (interquartile range, z-score, percentage change from the previous period), with their limits. Explain why automatic deletion is risky. Output a decision log template and a sample of three worked cases using my values.
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.