Home / Words / outlierNo. 0216

outlier

/ˈaʊtlaɪər/·noun

a point or person lying far outside the cluster the rest of the group forms
Fig. 1 — Eleven sheep graze shoulder to shoulder in the hollow, heads dipping almost in time; when the flock shifts a step, it shifts together.
01Definition

An outlier is the point that lies far outside the cluster the rest of the data makes — the reading much higher or lower than its neighbours, the town far richer or poorer than its region, the career no curve predicted. The word is four centuries old, but geology gave it its modern shape in the early nineteenth century, for rocks lying detached from their main formation — and it still means exactly that: something out-lying. An outlier is not automatically an error; it may be a slip of the instrument or the most interesting observation in the set, which is why analysts inspect one before deleting it.

02In use
  • iAnalysts flag any reading that deviates this far from the rest as an outlier before they average the set.
  • iiIn a valley of small farms, the vast estate on the hill is the outlier.
  • iiiHer career is an outlier among her classmates' — the same start, a wildly different arc.
03Collocations
  • an outlier in the data
  • a statistical outlier
  • flag outliers
  • remove outliers
  • an outlier among

Family outliers (plural noun) · outlying (adjective)

04Relations

=anomaly, exception, aberration, deviation, extreme

≠norm, average

06TOEFL & IELTS

IELTS Task 1 gold for the figure that breaks the pattern: 'the 2020 spike is an outlier' handles the awkward data point in one professional stroke, with 'a statistical outlier' and 'an outlier in the data' as the standing collocations. In Task 2 and speaking it stretches to people and cases — an outlier among its neighbours. Register is neutral-formal, safe anywhere academic. Pronounce it OWT-ly-er, three syllables; the adjective for places is outlying (outlying villages).

07Asked
What is an outlier in statistics?
A data point that differs markedly from the rest of the set — far above or below the cluster the other values form, like the lone sheep up the slope in the scene above. Statisticians commonly flag values more than about 1.5 times the interquartile range beyond the quartiles, or two to three standard deviations from the mean — and a single one can drag an average badly off course.
What is the difference between an outlier and an anomaly?
An outlier is a value lying far from the rest of its group; an anomaly is anything that breaks the expected pattern, whether or not it sits far away numerically. A perfectly average-looking transaction at 3 a.m. can be an anomaly without being an outlier. In practice data scientists use the terms loosely, but the instinct holds: outlier measures distance, anomaly measures wrongness.
What does it mean to call a person an outlier?
That they sit far outside the norm of their group — in results, behaviour, or path. The tone is set by context: admiring ('an outlier who rewrote the record book'), neutral ('an outlier in the survey'), or quietly dismissive ('that result is an outlier' — meaning don't build a theory on it). The word judges distance from the group, not merit; the direction of the distance is yours to supply.
Why is Malcolm Gladwell's book called Outliers?
Because its subjects — the Beatles, Bill Gates, elite athletes — are people whose success sits as far from the norm as a stray data point sits from its cluster. Gladwell's 2008 argument is that such people are not lone geniuses: he credits timing, opportunity, cultural legacy, and roughly ten thousand hours of practice. The book did more than anything else to move the statistician's word into everyday English.
Is outlier formal enough for an academic essay?
Yes — and it quietly signals quantitative thinking, which is why examiners like it. Its rivals carry different flavours: an exception implies a rule being broken, a special case sounds like pleading, while an outlier implies you have looked at the whole distribution and found one point standing apart. It also takes hedges comfortably — 'appears to be an outlier', 'may simply be an outlier' — letting you flag a strange result without dismissing it.
Should outliers be removed from data?
Only after inspection, and often not at all. Remove a value just for being extreme and you may be deleting the discovery — rare events, new phenomena, the one patient the drug harmed. Keep it blindly and a single wild value can drag the average badly off. The working rule: identify outliers mechanically, then decide their fate by cause — error out, evidence in.