1Describing a pile of numbers: centre and spread
Hand someone forty exam scores, a season of daily temperatures or a month of website response times, and their first need is compression: a few honest numbers that stand in for the pile. Descriptive statistics is that compression, and it always comes in two halves that must travel together. The centre says where the data lives (mean, median, mode). The spread says how scattered it is around that home (range, variance, standard deviation, interquartile range). A centre quoted without a spread is half a story — “average wait: 4 minutes” means one thing if every wait was 3–5 minutes and something entirely different if half were instant and half were 8.
Our calculator computes the full set in one pass and, just as importantly, shows the step-by-step working — the sorted list, the squared deviations, the quartile positions — because these statistics are where homework, lab reports and business dashboards all meet. The sections below take each measure in turn: what it captures, where it breaks, and which one to quote for which job.
One promise before we start: every statistic in this guide is computed on the same example dataset wherever possible, so you can watch the measures agree and disagree on identical evidence. That dataset appears in full in Section 7.
2Mean, median, mode: three kinds of “typical”
The mean is the arithmetic average — sum ÷ count — the balance point of the data. It uses every value, which is its strength and its weakness: one extreme value drags it. The median is the middle value of the sorted list (or the average of the two middle values for an even count). It ignores magnitudes at the extremes entirely — only positions matter — which makes it stubbornly resistant to outliers. The mode is the most frequent value; a dataset can have one, several (bimodal data is a clue that two groups are hiding inside it) or none.
The canonical illustration is income. In a room of nine people earning $40k and one earning $640k, the mean income is $100k — a number true of nobody in the room. The median is $40k, which describes nine of the ten perfectly. This is why official statistics quote median income and why data journalists reach for the median whenever money, house prices or anything with a long upper tail is involved. Conversely, the mean is the right choice when totals matter — mean rainfall × days = total rainfall — and the mode owns categorical data, where “average” is meaningless: the modal shoe size is a real, stockable thing.
Rule of thumb worth memorising: if the mean and median sit close together, the data is roughly symmetric and either will do. If they’ve drifted apart, the gap itself is information — something in the data is pulling, and Section 6 is about exactly that.
3Standard deviation: the honest measure of spread
The range (max − min) is a spread measure built from exactly two values — the two weirdest — so statisticians built a better one. The standard deviation asks: how far is a typical value from the mean? The recipe: subtract the mean from every value (the deviations), square them (killing the signs and punishing big misses), average the squares (that’s the variance), then take the square root to return to the original units. The squaring-then-rooting round trip is why students find it odd; the payoff is a measure in the data’s own units — seconds, marks, millimetres — that responds to every value in the set.
Its superpower is the empirical rule for bell-shaped data: about 68% of values fall within one standard deviation of the mean, 95% within two, 99.7% within three. Test scores with mean 70 and SD 10: roughly 95% of students scored between 50 and 90. Quality engineers live on this rule (a process whose defects sit beyond 3σ is a crisis; “six sigma” is a quality programme, not a slogan), and z-scores — (value − mean) ÷ SD — let you compare across different scales entirely: a score 1.5 SD above the mean is equally impressive in maths and in spelling.
The calculator shows both flavours of standard deviation — population (σ) and sample (s) — which differ by one small division and one big idea. That idea gets its own section, because it’s the most-missed point in introductory statistics.
4The N−1 story: population vs sample
You rarely measure every member of the group you care about. You measure a sample — 50 voters out of an electorate, 30 parts off a production line, one class out of a school — and use it to estimate the whole population. Here’s the subtle trap: a sample’s own mean is the best estimate of the population mean, but deviations measured around the sample mean are systematically a bit too small, because the sample mean was computed from those very values and hugs them closer than the true population mean would. Variance computed the plain way (÷ N) therefore underestimates the population’s variance.
The correction, called Bessel’s correction, is to divide by N−1 instead of N when computing sample variance. Dividing by a smaller number inflates the result by exactly the right amount, on average, to cancel the bias. That’s the entire difference between the two standard deviations your calculator reports: σ divides by N (use it when your data is the whole population — all 30 students in the class, all 12 months of the year), and s divides by N−1 (use it when your data is a sample standing in for something bigger).
How much does it matter? For N = 100, barely — the factor is 100/99. For N = 8, the sample SD is about 7% larger than the population SD, and choosing the wrong one in a lab report or an exam is the difference between matching the mark scheme and not. When in doubt, ask the question the statistic is answering: describing these values → population. Estimating a bigger group from these values → sample. Spreadsheet users meet this as STDEV.P vs STDEV.S — same fork, same logic.
5Quartiles and the IQR: the box plot’s skeleton
The median splits sorted data in half. The quartiles split it in quarters: Q1 is the median of the lower half (the 25th percentile — a quarter of the data sits below it), Q3 the median of the upper half (75th percentile). The distance between them, IQR = Q3 − Q1, is the spread of the middle 50% of your data — a robust spread measure that, like the median, ignores the extremes completely.
These five numbers (min, Q1, median, Q3, max — the “five-number summary”) draw the box plot, statistics’ most information-dense picture: a box spanning Q1 to Q3 with a line at the median, whiskers to the extremes. Box plots exist to compare groups at a glance — two classes’ scores, two factories’ part sizes — showing centre, spread and skew simultaneously in a way two means never could.
The IQR also powers the standard outlier rule: values below Q1 − 1.5 × IQR or above Q3 + 1.5 × IQR get flagged as outliers. It’s a convention, not a law of physics, but it’s the convention built into box-plot software everywhere — and our calculator reports the quartiles alongside the mean-based measures precisely so you can cross-examine your data from both the robust angle and the classical one. One honest caveat: quartile definitions vary slightly between textbooks and software for small datasets (whether the median is included in the halves). For large datasets the methods converge; for homework, match your teacher’s convention — ours uses the common median-of-halves (Tukey) method.
6Skew and outliers: when summaries lie
Data is skewed when one tail stretches further than the other. Right-skewed data (a long tail of big values — incomes, house prices, website session lengths) drags the mean above the median; left-skewed data (easy exam scores bunching at the top with a tail of low scores) drags it below. The mean–median gap from Section 2 is your skew detector, and it matters because every summary measure quietly assumes the shape of the data suits it. Quoting the mean of a heavily skewed set isn’t a rounding error; it’s a misdescription.
Outliers — genuinely extreme values — deserve investigation before deletion. Sometimes they’re errors (a temperature of 980° in a weather log is a sensor fault or a typo for 98.0). Sometimes they’re the most important values in the set: the one day the server fell over, the one batch that failed, the fraud hiding among the transactions. The honest workflow: flag with the IQR rule, investigate the cause, report results with and without if they were removed, and never delete silently. The median and IQR, being outlier-proof, are the summary of choice whenever the extremes are suspicious but must stay in the data.
Our calculator’s working box is designed for this examination — sorted values make skew visible, the squared-deviation table shows exactly which points contributed the variance, and quartiles sit next to the standard deviation so the two views of spread can be compared on the spot.
7One dataset, summarised five ways
Take the classic textbook dataset: 2, 4, 4, 4, 5, 5, 7, 9 (N = 8). Sum = 40, so the mean is 5. Sorted, the middle pair is 4 and 5, so the median is 4.5. The mode is 4 (three appearances). Range = 9 − 2 = 7. Deviations from 5: −3, −1, −1, −1, 0, 0, 2, 4; squares: 9, 1, 1, 1, 0, 0, 4, 16; sum of squares = 32. Population variance = 32 ÷ 8 = 4, so σ = 2. Sample variance = 32 ÷ 7 ≈ 4.571, so s ≈ 2.14. Quartiles: lower half {2, 4, 4, 4} → Q1 = 4; upper half {5, 5, 7, 9} → Q3 = 6; IQR = 2.
Read the summary as a paragraph: “Scores centred around 4.5–5, most commonly 4, typically within about 2 points of the mean, with the middle half packed tightly between 4 and 6 and a mild right pull from the 9.” Every number earned its place, and the N vs N−1 choice changed the SD by 7% on this tiny dataset — visible, exactly as Section 4 promised. Run the dataset through the calculator and every intermediate step above — the sorted list, each squared deviation — is printed for your working.
8Frequently asked questions
When should I use the median instead of the mean?
When the data is skewed or contains outliers you can't remove — incomes, house prices, response times with occasional huge spikes. The median describes the typical case honestly where the mean gets dragged. If mean and median are close, the data is roughly symmetric and either works; report both when you can.
What is the difference between σ and s on the calculator?
σ (sigma) is the population standard deviation — it divides by N and describes the dataset itself as the complete group. s is the sample standard deviation — it divides by N−1 (Bessel's correction) and estimates a larger population's spread from your sample. Exam papers and lab reports usually specify which; Excel's STDEV.P and STDEV.S are the same pair.
Why do quartiles differ between websites?
Because there are several legitimate quartile definitions (whether the median joins the halves, how positions interpolate) and small datasets expose the differences. For large datasets they all agree to rounding. Our calculator uses Tukey's median-of-halves method, the most common textbook convention — if your course uses another, your teacher's examples will reveal it.
Can standard deviation be negative or zero?
Never negative — it's a square root of an average of squares. Zero happens exactly when every value is identical (no spread at all), which is itself a useful diagnostic: an SD of 0 in measured data usually means the sensor is stuck or the same value was pasted down a column.
What does a z-score tell me?
How many standard deviations a value sits from the mean: z = (value − mean) ÷ SD. A z of +2 is unusually high (about the 97.7th percentile in bell-shaped data); −1 is unremarkable. Z-scores make different scales comparable — the foundation of standardised testing and of spotting which measurement in a mixed dashboard is truly abnormal.
How many numbers do I need for statistics to mean anything?
The calculator will summarise any count, but stability grows with N: means and SDs from 5 values wobble wildly if one value changes; by 30 they settle noticeably. There's no magic threshold — but treat small-sample statistics as provisional, and be doubly careful with the sample (N−1) formulas there, where the correction is largest.
Keep exploring
Statistics is one branch of the number family. Its neighbours:
Summarise any dataset with the working shown
Mean, median, mode, range, variance, both standard deviations and quartiles — every step printed, so you can learn it or check it. Free, private, no signup.
📊 Open the Statistics Calculator100% free · No account · Private by design
Related guides
More step-by-step guides on this topic:
