On this concept: Formula

Not sure which lesson to do first? Find your level and get a plan →

Math · Statistics & Probability · Grade 6-8

Outliers (Deep)

An outlier is a value lying unusually far from the bulk of the data, flagged when it falls below $Q_1-1.

Outlier if x<Q1−1.5×IQR or x>Q3+1.5×IQR

Quick Answer

An outlier is a value lying unusually far from the bulk of the data, flagged when it falls below Q1−1.5×IQR or above Q3+1.5×IQR. Use the rule when a value looks far from the rest and you need an objective way to decide. Find the quartiles, multiply the IQR by 1.5, and set a fence that far beyond each quartile; only values past a fence are flagged. Then decide what the outlier is: a typing error, a rare event, or an important exception worth studying.

Why This Matters

One outlier can wreck a summary: the incomes $50K, $55K, $60K, $58K and $5M have a mean of over $1 million, which describes nobody. Spotting outliers tells you when to switch from mean to median, which values to double-check for errors, and when an unusual reading is the most interesting thing in the data.

Intuitive Explanation

Imagine a class's 100 m times: most students finish between 14 and 20 seconds, and one time is written down as 41 seconds. You can see it sits apart. But is 21 seconds unusual too? Or 23? Judging by eye gives different answers for different people, so statisticians use a fixed rule built from the quartiles.

The quartiles mark the middle half of the data: Q1 is a quarter of the way up the sorted list and Q3 is three quarters of the way up. Their gap, the IQR, measures how wide the typical middle is. The fences go 1.5 IQRs beyond each quartile. If the middle half runs from 15 to 19 seconds, the IQR is 4, the step is 1.5×4=6, and the fences sit at 15−6=9 and 19+6=25 seconds. The 41 is flagged; 21 and 23 are not.

Being the largest value is not the same as being an outlier. Every data set has a maximum, and usually it is perfectly ordinary. An outlier is also different from noise: noise is the small jitter in every value, while an outlier is one value far from the rest. Once a value is flagged, find out why. A 41-second time might be a stopwatch error or a student who twisted an ankle, and those lead to different decisions about keeping it.

Think of it like this. Picture a teacher watching a class on a field trip. The group drifts along as a cluster, and the teacher keeps an imaginary boundary some distance beyond its edges, wider when the group is already spread out. A student a few steps from the others is fine; one beyond the boundary gets checked on, not sent home without a question.

Core idea. An outlier sits beyond 1.5×IQR past the quartiles — unusually far from the rest of the data.


When to Use

Use the fence rule when one or two values look far from the rest, when a mean seems dragged by an extreme value, or when a box plot asks you to mark outliers as separate dots. You need the quartiles first. If you only want the largest or smallest value, that is the maximum or minimum; if you want the total extent of the data, that is the range, which an outlier inflates rather than detects.

Tip. Ask: Does this value fall beyond Q1−1.5×IQR or Q3+1.5×IQR?

How to Recognize It

Before you call a value an outlier, find the quartiles of the whole data set, because the fence depends on how spread out the middle half already is.

  1. Is the question just asking for the largest or smallest value?

    If yes, that is the maximum or minimum, and nothing needs testing. Every data set has a largest value, but it counts as an outlier only if it lies past the fence.

  2. Does the suspicious value sit far from the rest while the other values stay close together?

    If yes, compute the fences and test it. If many values wobble a little with no single one standing apart, that scatter is noise, and there is no single point to flag.

  3. Does the question want how much the data covers from end to end?

    Then it wants the range, maximum minus minimum. An outlier makes the range larger, so the range shows the outlier's effect but cannot tell you whether a value is unusual.

  4. Do you know how the value was recorded, such as typed into a survey or read by a sensor?

    If yes, use that to decide what to do with a flagged value: a height of 1,500 cm is a typing slip and can be corrected, while a real record-breaking time stays in and gets reported.

Outliers (Deep) compared with Maximum / minimum, Noise, and Range

Outliers, the maximum and minimum, the range and noise all involve values at the edges or off the pattern, so the biggest number often gets called an outlier by default. Ask whether you are naming an extreme, measuring total width, describing small scatter, or testing a value against a fence.

Outliers (Deep)

Meaning
Use this when an extreme value must be judged unusually far from the rest using an objective rule.
Key test
Does this value fall beyond Q1−1.5×IQR or Q3+1.5×IQR?
Formula
Outlier if x<Q1−1.5×IQR or x>Q3+1.5×IQR
Example
Data has Q1=20, Q3=40, and a value of 75. Is 75 an outlier?

Maximum / minimum

Meaning
Is simply the largest or smallest value, which may be perfectly typical.
Key test
Use when you just want the extreme value, not whether it's unusual.
Example
Tallest student at 175 cm, still normal

Noise

Meaning
Is small random fluctuation across many points, not one far-off value.
Key test
Use when variation is widespread and minor, not a single spike.
Example
Day-to-day step-count wiggle

Range

Meaning
Measures total extent and is inflated by outliers, not a test for them.
Key test
Use when you want the spread number, then check it for outlier inflation.
Formula
max−min
Example
Range jumps from 20 to 200 due to one point

Formula & Notation

Outlier if x<Q1−1.5×IQR or x>Q3+1.5×IQR
x is an outlier if x<Q1−1.5⋅IQR or x>Q3+1.5⋅IQR where IQR=Q3−Q1

How to read it: Values beyond 1.5×IQR from the quartiles are called outliers; beyond 3×IQR are extreme outliers

Worked Examples

Example 1 — Flagging an outlier

Easy

Data has Q1=20, Q3=40, and a value of 75. Is 75 an outlier?
Show the steps and the answer
  1. Find the IQR: 40−20=20.

    The fences are measured in IQRs, so the width of the middle half comes first.

  2. Find the step: 1.5×20=30.

    The rule allows values up to 1.5 IQRs beyond a quartile before flagging them.

  3. Build the fences: upper 40+30=70, lower 20−30=−10.

    Since 75 is a large value only the upper fence matters, but both come from the same step.

  4. Compare: 75>70.

    A value past either fence is an outlier by this rule, and 75 is 5 past the upper one.

Answer: Yes — 75 is an outlier

Takeaway: Compute the IQR, then 1.5 times it, then the fences, and only then compare. Here 75 clears the fence of 70.

Example 2 — Just the maximum

Standard

Same data Q1=20, Q3=40, but the largest value is 65. Outlier?
Show the steps and the answer
  1. Reuse the fence: with the same quartiles, the upper fence is still 40+1.5×20=70.

    The fence depends only on the quartiles, not on the value you are testing.

  2. Compare: 65<70.

    65 lies inside the fence, within the normal reach of this data.

  3. Note what being the maximum means: some value must be the largest, and here it is 65.

    Every data set has a maximum, so being the biggest value is no evidence of being unusual.

Answer: No — 65 is below the fence 70, so not an outlier

Takeaway: The largest value is an outlier only if it passes the fence. 65 is the maximum but stays 5 inside it.

Example 3 — Find the mistake: judging by the gap

Application

Eleven students' homework times, in minutes, are 15,20,22,25,25,28,30,32,35,40,52. Jo writes: "52 is 12 minutes above the next-highest time, so it is an outlier." Find the mistake and correct it.
Show the steps and the answer
  1. Find the quartiles: the median is the 6th value, 28; the lower five values 15,20,22,25,25 give Q1=22; the upper five 30,32,35,40,52 give Q3=35.

    The rule is built from the quartiles, not from the gap between two neighbouring values.

  2. Compute IQR=35−22=13 and the step 1.5×13=19.5.

    The step grows with how spread out the middle of this data already is.

  3. Build the upper fence: 35+19.5=54.5 minutes.

    Only values above 54.5 count as unusually high for this data.

  4. Compare: 52<54.5, so 52 is not an outlier.

    Jo judged by one gap; the middle half of this data is already 13 minutes wide, so a 12-minute gap is not unusual here.

Answer: 52 is not an outlier: the upper fence is 35+1.5×13=54.5, and 52 is below it.

Takeaway: A big gap to the next value is not the test. Build the fence from the quartiles and compare.

Where You'll See It

  • Catching typos in a spreadsheet

    A school records students' heights in centimetres, with Q1=150 and Q3=165. The IQR is 15, so the fences are 150−22.5=127.5 and 165+22.5=187.5. An entry of 1580 is flagged at once; it is almost certainly 158.0 typed without its decimal point.

  • Spotting credit card fraud

    A bank knows a customer's purchases usually run from $10 to $80. A $2,400 charge in another country lands far beyond the fence for that customer, so the bank freezes the card and texts to ask whether it was really them. Here the outlier is not deleted: it is the whole point of looking.

Common Mistakes

  1. The slip: Eyeballing outliers without the fence

    Instead: compute Q1−1.5×IQR and Q3+1.5×IQR to flag them.

  2. The slip: Deleting outliers automatically

    Instead: first decide if it's an error, a rare event, or a meaningful exception.

  3. The slip: Calling the maximum an outlier by default

    Instead: the largest value isn't unusual unless it passes the fence.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A data set has Q1=30 and Q3=50. Find the lower and upper outlier fences.

    Hint: Find the IQR, multiply by 1.5, then step out from each quartile.

  2. Is 25 an outlier in 2,4,5,6,7,8,9,25?

    Hint: Find Q1 and Q3 from the two halves of 4 values.

  3. For 3,12,13,14,15,16,18,22, check whether 3 and 22 are outliers.

    Hint: Find both fences, then compare each value with them.

  4. Test scores are 80,82,85,88 and 15. Find the mean with and without the 15.

    Hint: Find each total, then divide by the right count.

  5. A room thermometer logs 21,22,21,23,22,85 degrees Celsius. Is the 85 noise or an outlier, and what should you do?

    Hint: Is it a small wiggle like the others, or one far-off value?

  6. For 4,6,7,8,9,10,11,13, Jin says: "13 is an outlier because it is the largest value." Correct him.

    Hint: Compute the upper fence before deciding.

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

Why do waiting times almost never have low outliers?

Waiting times cannot go below zero, and they usually have a long tail to the right. If Q1=4Q_1 = 4 and Q3=12Q_3 = 12 minutes, the IQR is 8 and 1.5×IQR=121.5\times\text{IQR} = 12, so the lower fence is 4−12=−84 - 12 = -8, a wait no one can have. The upper fence, 12+12=2412 + 12 = 24, is easy to pass. In data with a floor at zero and a long right tail, the flagged values almost always come from the top.

After I remove an outlier, should I check again for new ones?

Usually not. Removing a value changes the quartiles, and the new, narrower fences may flag values that looked fine before, so repeating the test can eat away at real data one point at a time. Test once on the full data, decide what each flagged value is, and report what you removed and why.

Has an unusual value ever led to a big discovery?

Yes. In 1985, scientists of the British Antarctic Survey reported spring ozone readings over Antarctica far below anything in their earlier records. They first suspected their instrument and checked it, but the low values were real. Taking those strange numbers seriously instead of throwing them out revealed the ozone hole, and in 1987 countries signed the Montreal Protocol to phase out the chemicals causing it.

Learning Path

Outliers (Deep)

You are here

Next →

Box Plot
You need Variability and the Interquartile Range first, because the quartiles and the IQR are what the fences are built from. This concept adds a fixed rule for flagging unusual values and the habit of asking why they are unusual. Next, the Box Plot draws the quartiles, whiskers out to the furthest values inside the fences, and each outlier as its own dot.

See Also