statcompute

Category

Descriptive Statistics: Summarising a Column of Numbers

Before any test, describe the data. These tools summarise a column of numbers and show you its shape, which is what decides which test is appropriate in the first place.

Centre, spread, shape — in that order

Every summary of a data set answers three questions. Where is it? — the mean, median and mode. How spread out is it? — the standard deviation, variance, range and mean absolute deviation, or the interquartile range. What shape is it? — skewness, kurtosis, and whether there are outliers.

The three interact. On symmetric data the mean and median agree and either will do. On skewed data they separate, and the gap is informative: mean income exceeds median income everywhere in the world, because a long right tail pulls the mean and leaves the median where it is.

Why the mean and the median can tell opposite stories

The mean uses every value's magnitude; the median uses only their order. That makes the mean efficient when the data are well behaved and misleading when they are not. One mistyped value — a weight of 700 kg instead of 70 — moves the mean of a sample of fifty by more than 12 kg and moves the median by nothing at all.

The same split runs through the spread measures. The standard deviation squares every deviation, so an outlier dominates it; the median absolute deviation does not. When a robust and a non-robust summary disagree sharply, that disagreement is a finding: something extreme is in the data.

Sample or population?

Nearly every formula here has two versions. The population version divides by n; the sample version divides by n − 1, because deviations measured from the sample mean are systematically too small. Use the sample version unless your data genuinely are the entire population — every employee, every transaction, every item — rather than a sample drawn from one. This site defaults to the sample version and shows both.

Then look at it

Numbers summarise; pictures reveal. Two data sets can share a mean, a standard deviation and a correlation and look nothing alike. The box plot, histogram and stem-and-leaf plot makers all show the distribution these numbers came from, and each calculator here draws one automatically.

Left-skewedmean below medianSymmetricmean = medianRight-skewedmean above mediandashed lines: mean and median
The mean chases the tail and the median does not. That is the whole reason to report both, and the reason median income is quoted rather than mean income.
01

The 17 calculators in this category

02

Questions people ask

Should I divide by n or n − 1?

By n − 1 when the data are a sample and you want to estimate the population variance — that is Bessel's correction, and it exists because deviations from the sample mean are too small on average. Divide by n only when the data are the whole population. Every variance and standard deviation result here reports both.

Why does my calculator give a different quartile from a spreadsheet?

Because there are at least four accepted definitions and they genuinely disagree on small samples. Excel's PERCENTILE.INC and R's default use linear interpolation; PERCENTILE.EXC and R's type 6 use a different position; box plots traditionally use Tukey's hinges. The percentile calculator shows all four side by side and names the one it used.

What counts as an outlier?

There is no universal answer — only rules. The box-plot rule flags anything beyond 1.5 × IQR from the quartiles; the z-score rule flags |z| > 3; the modified z-score uses the median and is the most reliable on small samples. The outlier calculator applies all three, and none of them tells you whether to delete the point. That needs a reason from outside the data.