Calculate lower and upper fences for outlier detection using IQR method with box plot visualization, outlier classification, and clean-data impact analysis.
The lower fence and upper fence are boundary values used in the IQR (Interquartile Range) method to identify outliers — data points that are unusually far from the bulk of the data. The lower fence equals Q1 − 1.5 × IQR and the upper fence equals Q3 + 1.5 × IQR. Any value below the lower fence or above the upper fence is classified as an outlier.
This calculator computes both fences using either inclusive or exclusive quartile methods, generates a complete box plot visualization, and classifies every data point into five zones: extreme low, mild low, normal, mild high, and extreme high. The impact analysis shows exactly how removing outliers changes the mean and standard deviation, helping you decide whether to remove or retain them.
The 1.5 × IQR rule was introduced by John Tukey and remains the most widely used outlier detection method in exploratory data analysis. This calculator also supports the 3 × IQR threshold for identifying only extreme outliers, which may represent data entry errors rather than legitimate extreme values.
The IQR fence method is the gold standard for outlier detection in exploratory data analysis because it's robust — outliers don't distort the detection boundaries the way they do with mean/SD-based methods. This calculator provides both the standard (1.5×) and extreme (3×) thresholds, five-zone classification, and a visual box plot.
The clean-data impact analysis is critical for decision-making: it shows exactly how much the mean and standard deviation change when outliers are removed, helping you assess whether outliers are materially affecting your analysis. The box plot visualization makes outlier positions immediately clear, even for non-statisticians.
IQR = Q3 − Q1 Lower Fence = Q1 − k × IQR Upper Fence = Q3 + k × IQR where k = 1.5 for mild outliers, k = 3 for extreme outliers Outlier classifications: - Extreme low: value < Q1 − 3 × IQR - Mild low: Q1 − 3 × IQR ≤ value < Q1 − 1.5 × IQR - Normal: Q1 − 1.5 × IQR ≤ value ≤ Q3 + 1.5 × IQR - Mild high: Q3 + 1.5 × IQR < value ≤ Q3 + 3 × IQR - Extreme high: value > Q3 + 3 × IQR
Result: Lower Fence = 30.25, Upper Fence = 76.75, IQR = 11.5, Upper Outlier: 150
Q1 = 47.50, Q3 = 59.00, IQR = 11.50. Lower fence = 47.50 − 17.25 = 30.25, upper fence = 59.00 + 17.25 = 76.75. The value 150 far exceeds the upper fence and is classified as an extreme outlier (>3×IQR). It could be a data entry error (salary of $150k among $43-72k values). Removing it changes the mean from $56.10 to $51.16 — a significant shift.
John Tukey introduced the box-and-whisker plot in his 1977 book "Exploratory Data Analysis." The fences at 1.5× and 3× IQR were designed to capture "outside" and "far outside" values respectively. Tukey's approach revolutionized statistics by emphasizing visual exploration over hypothesis testing, and the box plot remains the most widely used data visualization for distribution comparison.
The IQR method assumes unimodal, roughly symmetric distributions. For multimodal data, strongly skewed distributions, or very small samples (n < 10), the fences may be misleading. Adjusted box plots for skewed data use the medcouple (MC) to create asymmetric fences: Q1 − 1.5 × e^(−4MC) × IQR for the lower fence and Q3 + 1.5 × e^(3MC) × IQR for the upper fence when MC ≥ 0.
Beyond simple removal, analysts can: (1) Winsorize — replace outliers with the fence values; (2) Trim — compute trimmed means excluding the most extreme k%; (3) Transform — apply logarithmic or square root transformations to compress the range; (4) Use robust methods — median, IQR, and rank-based tests that are naturally resistant to outliers. The right approach depends on why the outlier exists and what analysis follows.
The minimum is the smallest actual data value. The lower fence is a calculated boundary (Q1 − 1.5×IQR). The lower fence can be below, at, or above the minimum. If the minimum is below the lower fence, it's an outlier; if above, there are no lower outliers.
John Tukey chose 1.5 empirically. For normally distributed data, the 1.5×IQR rule flags about 0.7% of data (roughly the same as ±2.7σ). The value 3 captures about 0.0002% (roughly ±4.7σ). These thresholds balance sensitivity with false positive rates for most practical distributions.
Not automatically. Outliers may be data entry errors (remove or correct them), legitimate extreme values (keep them), or signs of a different subpopulation (investigate). The impact analysis shows what happens if you remove them. If the change is dramatic, consider reporting both analyses.
The inclusive method includes the median in both halves when computing Q1 and Q3. The exclusive method excludes the median from both halves. They can give different Q1/Q3 values for odd-sized datasets. Tukey originally used the exclusive method, but many textbooks use inclusive.
Yes. If Q1 is small and IQR is large, the lower fence can be negative even when all data is positive. This simply means no lower outliers are possible in non-negative data. A negative lower fence doesn't indicate an error.
Z-scores flag values beyond ±2 or ±3 standard deviations. The problem is that outliers inflate the mean and SD, making the z-score method less sensitive (masking). The IQR method is robust because Q1, Q3, and IQR are not affected much by extreme values.