How Does Changing One Data Value Affect the Variance?
Change a single number in a data set and the variance can barely move or jump dramatically. Which one happens is not random: it depends on where the new value sits relative to the mean. Understanding this explains why one unusual measurement can distort an entire analysis, and why outliers deserve a closer look.
Worked Example: Moving One Value Away From the Mean
Start with five measurements: 10, 12, 11, 13, 14.
- The mean is 60 / 5 = 12.
- Deviations: −2, 0, −1, 1, 2. Squared: 4, 0, 1, 1, 4.
- The sum of squares is 10, so the population variance is 10 / 5 = 2.
Now suppose the last measurement is entered as 24 instead of 14. The data set becomes 10, 12, 11, 13, 24.
- The mean rises to 70 / 5 = 14.
- Deviations: −4, −2, −3, −1, 10. Squared: 16, 4, 9, 1, 100.
- The sum of squares is 130, so the population variance is 130 / 5 = 26.
One value moved by 10, and the variance went from 2 to 26, thirteen times as large. The new value alone contributes 100 of the 130, because its deviation of 10 is squared.
Worked Example: Moving One Value Towards the Mean
Go back to the original data and change 14 to 12 instead. The data set is now 10, 12, 11, 13, 12.
- The mean is 58 / 5 = 11.6.
- Deviations: −1.6, 0.4, −0.6, 1.4, 0.4. Squared: 2.56, 0.16, 0.36, 1.96, 0.16.
- The sum of squares is 5.2, so the population variance is 5.2 / 5 = 1.04.
Moving a value closer to the centre lowered the variance from 2 to 1.04.
| Change | Mean | Sum of squares | Population variance |
|---|---|---|---|
| Original data | 12 | 10 | 2 |
| 14 → 24 (away from the mean) | 14 | 130 | 26 |
| 14 → 12 (towards the mean) | 11.6 | 5.2 | 1.04 |
Why the Whole Calculation Changes
Every value is measured from the mean, and the mean depends on every value. So changing one number moves the mean, which changes every deviation, not just the one you edited. In the first example, the four untouched values all ended up further from the new mean of 14 than they were from 12.
Why Extreme Values Matter So Much
Because deviations are squared, distance counts more than proportionally. A value twice as far from the mean contributes four times as much; three times as far contributes nine times as much. This is why a single extreme value can dominate the total, and why statisticians check outliers before trusting a variance.
Adding or Removing a Value
Adding an observation changes n as well as the mean. If the new value is close to the mean, the variance usually falls slightly; if it is far away, the variance rises. Removing an extreme value often reduces the variance sharply, which is a quick way to see how much one point was driving the result. Report both figures if you are unsure whether the point is a genuine measurement.
Recalculating After an Edit
For small data sets, the safest approach is to recalculate from scratch: new mean, new deviations, new squares, new total. Paste the edited data into the variance calculator and compare the result with the original.
Summary
Changing one value shifts the mean and therefore every deviation. Values pushed away from the mean raise the variance sharply because distances are squared; values pulled towards it lower it. That sensitivity is a property of variance, not a flaw, and it is the reason to look closely at unusual values before relying on the number.