How this instrument works
An outlier, in the sense this instrument uses, is not just 'a big number' — it is a value that sits unusually far from the bulk of the data, measured against the data's own spread rather than against a fixed threshold. The most common rule for flagging one, due to statistician John Tukey, is the 1.5×IQR fence: find the interquartile range (Q3 − Q1, the middle 50% of the data), then draw a lower fence 1.5 IQRs below Q1 and an upper fence 1.5 IQRs above Q3. Anything outside those two fences is flagged as an outlier; everything inside is treated as ordinary variation.
The fence multiplier of 1.5 is a convention, not a law of nature — it was chosen because, for data that is roughly normally distributed, it flags only about 0.7% of points as outliers, a reasonably conservative rate. Some fields tighten it (3×IQR is sometimes used for 'extreme' outliers) or use other methods entirely (z-scores, modified z-scores with the median), but the 1.5×IQR fence is the one taught in most introductory statistics courses and used by default in most box-plot software, which is why it is the version built into this instrument.
This calculator's specific job is narrower than just computing the fences: you give it a data set and one candidate value, and it tells you directly whether that value falls outside the fences — a single yes/no answer, alongside the fence numbers themselves so you can see how close the call was.
- Paste your numbers into Data set — commas, spaces or new lines all work as separators.
- Enter the number you want checked into Value to test — it does not need to already be a member of the data set.
- Read Lower fence and Upper fence — anything between them is ordinary; anything outside is flagged.
- Read Is the test value an outlier? — it returns 1 if your test value falls outside either fence, 0 if it falls inside them.
- To screen an entire list rather than one suspect value, test each candidate value in turn, or scan by eye for anything below the lower fence or above the upper one.
Worked example — testing the highest score in a set of eight
Take the eight values {2, 4, 4, 4, 5, 5, 7, 9} — the same teaching data set used elsewhere on this site — and test whether the largest value, 9, is an outlier. By linear interpolation, Q1 = 4 and Q3 = 5.5, so IQR = 1.5. The lower fence sits at Q1 − 1.5×1.5 = 1.75 and the upper fence at Q3 + 1.5×1.5 = 7.75.
Enter this data set with Value to test set to 9, and the instrument returns Lower fence = 1.75, Upper fence = 7.75, and Is the test value an outlier? = 1. Nine is above 7.75, so it lands outside the upper fence — flagged as an outlier by this rule, even though it is not an extreme value in absolute terms.
Questions
What does the 1=yes / 0=no output actually mean?
It is the instrument's direct answer to 'is my test value an outlier': 1 means the value you entered falls outside the lower or upper fence and is flagged; 0 means it falls between the fences and is treated as ordinary variation. It only evaluates the single value you typed into Value to test, not every point in your data set at once.
Why 1.5×IQR specifically, and not some other multiplier?
It's John Tukey's original convention from box-plot analysis, chosen because for roughly normal data it flags only about 0.7% of points, a conservative rate that catches genuinely unusual values without over-flagging ordinary spread. Some analysts use 3×IQR for a stricter 'extreme outlier' cutoff, but 1.5×IQR is the version taught in most statistics courses and used by default in most box-plot software, which is why it's the rule built into this instrument.
Can I test a value that isn't part of my original data set?
Yes — Value to test is independent of Data set. You could enter a new measurement you're deciding whether to trust, or a competitor's reported figure, and check it against the fences computed from your existing data, without adding it to the data set first.
I just want the fence numbers, not a yes/no verdict — is there a simpler tool?
If you're building a box plot by hand or just want the boundary values themselves, this site's dedicated lower-fence, upper-fence and combined upper-lower-fence calculators report exactly those numbers without requiring a test value. This tool is built for the narrower question of screening one specific candidate.
Does being outside the fences prove a value is an error?
No — the 1.5×IQR rule flags statistical unusualness, not correctness. A flagged value might be a data-entry mistake, or it might be a real, valid extreme measurement (a genuinely exceptional exam score, a real spike in a sensor reading). The rule tells you where to look twice; deciding whether to exclude the point is a judgment call that depends on domain knowledge the formula doesn't have.
What's the smallest data set this works on?
The quartile calculation needs at least a handful of points to be meaningful — two or three values give a technically computable but not very informative IQR, since the middle 50% barely exists. In practice this rule is most useful and most commonly applied to data sets of at least 5–10 points.