Last updated: 2026-07-13

Outliers and the IQR method

An outlier is a value that sits far away from the rest of a data set. Outliers can be genuine — a record sales day, an unusually tall person — or they can be mistakes, such as a mistyped digit or a faulty sensor reading. Either way, they distort averages and inflate measures of spread, so it helps to have a consistent rule for spotting them.

The interquartile range (IQR) method, introduced by the statistician John Tukey, is one of the most widely used rules. Instead of relying on the mean and standard deviation — which are themselves pulled around by extreme values — it builds its boundaries from quartiles, which are far more stable. That makes the method a reliable default, especially for skewed data.

Quartiles and the interquartile range

Sort the data and split it into four equal parts. The cut points are the quartiles:

• Q1, the first quartile, is the 25th percentile: a quarter of the values fall below it.
• Q2 is the median, the 50th percentile.
• Q3, the third quartile, is the 75th percentile: three quarters of the values fall below it.

The interquartile range is the distance between the outer two:

IQR = Q3 − Q1

It measures the spread of the middle half of the data and ignores the tails entirely, which is exactly why it resists the influence of outliers.

The 1.5×IQR rule

The method places two boundaries, called fences, on either side of the central box:

• Lower fence = Q1 − 1.5 × IQR
• Upper fence = Q3 + 1.5 × IQR

Any value below the lower fence or above the upper fence is flagged as an outlier. The multiplier 1.5 is Tukey's convention: for normally distributed data it sits about 2.7 standard deviations from the mean, so only around 0.7% of values are expected to fall outside the fences by chance. A smaller multiplier would flag too many ordinary points; a larger one would miss real anomalies.

Worked example

Take the data set 10, 12, 12, 13, 12, 11, 40, 12, 11. Sorted, it becomes:

10, 11, 11, 12, 12, 12, 12, 13, 40

There are n = 9 values. Using linear interpolation between order statistics:

• Q1 (25th percentile) = 11
• Q3 (75th percentile) = 12
• IQR = 12 − 11 = 1

The fences follow directly:

• Lower fence = 11 − 1.5 × 1 = 9.5
• Upper fence = 12 + 1.5 × 1 = 13.5

Now compare each value against the fences. Everything from 10 to 13 lies comfortably inside, but 40 is far above the upper fence of 13.5. So 40 is the single outlier in this set — which matches the intuition that one value clearly does not belong with the rest.

Mild versus extreme outliers

The rule is often applied at two levels. Values beyond the 1.5×IQR fences are mild outliers. Widening the fences to 3.0×IQR gives extreme outliers:

• Lower extreme fence = Q1 − 3.0 × IQR
• Upper extreme fence = Q3 + 3.0 × IQR

Every extreme outlier is also a mild one, but not the other way around. The distinction is useful for triage: a mild outlier is often just a naturally large or small observation, while an extreme outlier is far enough out that a data-entry error or a genuinely rare event becomes more plausible. This calculator uses the standard 1.5×IQR fences; to look for extreme outliers, recompute the fences with a multiplier of 3.0.

The connection to box plots

The IQR method is the mathematics behind the box plot. The box spans Q1 to Q3, a line inside marks the median, and the whiskers extend to the most distant values that still lie within the 1.5×IQR fences. Any point past the whiskers is drawn individually as an outlier. Reading a box plot is therefore a visual version of exactly this calculation.

What to do with an outlier

Detecting an outlier is the beginning of an investigation, not an instruction to delete it. Start by checking whether the value is an error — a wrong unit, a slipped decimal point, a duplicated entry — and fix it if so. If the value is real, it may carry the most interesting information in the whole data set, and discarding it would hide something worth understanding.

When outliers are kept, robust summaries such as the median and the IQR describe the data more faithfully than the mean and standard deviation, which a single extreme value can distort. Any decision to exclude a point should be recorded and justified, never made quietly.

Related calculators

• Percentile and Quartile Calculator — quartiles and percentiles for a data set
• Mean, Median, Mode Calculator — measures of center, including the robust median
• Variance and Standard Deviation Calculator — spread measured from the mean
• Descriptive Statistics Calculator — a full summary of a data set