How this instrument works
Think of a packaging line logging fill weights all shift, or a batch of lab results coming back from a set of patients or samples. Almost everything lands in a tight cluster, a few readings trail off toward the low end, and now and then one comes in so far below the rest that it looks like something went wrong upstream rather than the process just running a bit light. The lower fence turns that suspicion into an actual number: lower fence = Q1 − 1.5 × IQR, built from Q1, the first quartile, stepped back by 1.5 interquartile ranges. Drop beneath that line and a reading has left the territory of ordinary low variation.
That 1.5 figure isn't a physical constant — it's a convention exploratory data analysis adopted because it works: forgiving enough that a routinely lighter-filling nozzle or a naturally quieter patient reading won't set off the alarm, strict enough to still catch a reading that has genuinely fallen away from the pack. Some quality programs tighten the multiplier to 3 when they want to flag only the most severe shortfalls rather than every merely-possible one.
What comes back here is the boundary itself, which is exactly what you need before you know which specific reading you'll be checking — setting a documented reject threshold on a filling line, sketching the lower whisker of a box plot, or scanning a whole batch of results for anything suspiciously low in one pass. Got one particular reading you already suspect? This site's outlier calculator takes that single value alongside your data and returns a direct yes-or-no instead of leaving you to do the comparison by eye.
- Paste or type your numbers into the Data set field — commas, spaces, or new lines all work as separators.
- The instrument sorts your data and finds Q1 and Q3 by linear interpolation, then computes the interquartile range between them.
- Read the result in Lower fence — any value in your data (or any new value you are screening) below this number counts as a low-side outlier by the standard 1.5 × IQR rule.
- Use it as the bottom boundary when hand-drawing a box plot's lower whisker, or as a documented low-side screening threshold for new data.
- Needs at least 2 values to compute quartiles — for a specific yes/no check on one number, use this site's outlier calculator instead.
Worked example — the eight quiz scores
Enter {2, 4, 4, 4, 5, 5, 7, 9} in Data set — eight quiz scores, n = 8. Interpolating the sorted list gives Q1 = 4 and Q3 = 5.5, a 1.5-point IQR. Take 1.5 times that IQR away from Q1: 4 − (1.5 × 1.5) = 1.75, which is what Lower fence shows.
This class's weakest score, a 2, stays above that 1.75 cutoff — it's the smallest number on the roster, but it never actually crosses into low-outlier territory. That gap matters: being the bottom score in a group and being a statistical anomaly are two separate claims, and only the second one is what the lower fence is built to test for.
Questions
What does crossing the lower fence actually indicate?
A reading below Q1 − 1.5 × IQR has dropped further from the pack than routine low-end variation would explain. It's flagged as worth a second look — a possible sensor fault, a shortfall, a data-entry slip — not automatically thrown out or assumed wrong.
Where does the 1.5 multiplier in the formula come from?
It's a convention out of exploratory data analysis, picked because it holds up in practice: forgiving enough to leave ordinary quiet readings alone, tight enough to still catch ones that have genuinely fallen away. Some processes tighten it to 3 when only the most severe shortfalls should trigger a flag, leaving 1.5 for the everyday 'worth checking' cases.
My lower fence came back negative — is that a mistake?
No, that happens whenever Q1 is smaller than 1.5 × IQR, and it's a perfectly valid result. For data that can never physically go negative — counts, weights, durations — a negative lower fence effectively means there's no real low-side cutoff at all, since nothing in that data could ever fall past it.
I have one specific reading I'm suspicious of — how do I test just that one?
Compare it against the Lower fence figure directly, or feed it into this site's outlier calculator, which accepts a single test value alongside your full data set and hands back a straight yes-or-no rather than requiring you to eyeball where it lands.
Should I also check the upper side, or does the lower fence cover enough on its own?
That depends on what you're worried about. Chasing only shortfalls, dropouts, or unusually small readings? The lower fence alone does the job. Need both boundaries — for a complete box plot, or a screen that watches for high and low problems together — grab both from this site's upper/lower fence calculator in one step.
If a value falls below the fence, do I have to exclude it from my analysis?
Not automatically. The fence only says a reading is statistically unusual relative to everything else in your data — it doesn't diagnose why. Deciding whether to investigate, correct, keep, or drop it takes context the formula alone can't provide, like whether it traces back to an equipment fault or is simply a genuine, valid low reading.