Client-Side UtilityLocal memory execution • Zero server telemetry
Outlier Finder — Detect Anomalies via IQR & Z-Score
Identify statistical outliers, anomalous data points, and extreme deviations in your dataset using Tukey’s Interquartile Range (IQR) fences and Gaussian Z-Score standards.
DMaintained by Data & Statistics Desk
Published
Detection Methodology
Anomaly Analysis Results
Enter at least 4 numerical values on the left to identify statistical outliers.
DMaintained by Data & Statistics Desk
Published
Calculation Methodology & Formula
Tukey IQR method detects values outside Inner Fences [Q1 - 1.5×IQR, Q3 + 1.5×IQR] and Outer Fences [Q1 - 3×IQR, Q3 + 3×IQR]. Z-Score method flags observations where |x - μ| / σ > threshold.
\text{Fence}_{\text{inner}} = [Q_1 - 1.5\cdot\text{IQR}, \; Q_3 + 1.5\cdot\text{IQR}] \quad \Big| \quad Z = \frac{x - \bar{x}}{s}Worked Example: Lab Experiment Anomaly Detection
Identify measurement anomalies in sensor readings: 10, 11, 12, 10, 11, 13, 9, 85, 12.
dataset:10, 11, 12, 10, 11, 13, 9, 85, 12
method:Tukey's IQR Fences
Expected OutputOutlier: 85 (Extreme High). Upper Inner Fence: 16.5.
85 lies far beyond Q3 + 3×IQR (21.0), classifying it as an extreme statistical anomaly.
Technical Constraints & Limitations
- Tukey’s IQR method requires at least 4 numerical values to compute meaningful quartiles.
- Z-Score method assumes an approximately normal distribution; datasets with heavy skew may flag false positives.
Frequently Asked Questions
What is the difference between Mild and Extreme outliers?
In Tukey’s boxplot rule, Mild outliers lie between 1.5×IQR and 3.0×IQR away from the nearest quartile. Extreme outliers lie more than 3.0×IQR away and often signify severe measurement errors or black-swan events.
Should I choose IQR or Z-Score?
Use IQR for skewed, financial, or non-normal data because it is not distorted by extreme values. Use Z-Score for normally distributed Gaussian datasets (e.g., standard testing scores or sensor calibration).
Can I export clean data with outliers removed?
Yes. You can use the "Clean Dataset (Inliers Only)" download action to export the sanitized data list instantly.