Descriptive statistics are the numbers that summarise a dataset without trying to draw a conclusion from it. This descriptive statistics calculator produces the whole standard set in a single pass: count, sum, minimum, maximum, range, mean, median, mode, sample and population variance, standard deviation, standard error, the three quartiles, the interquartile range, and the two shape coefficients — skewness and kurtosis — that describe how the distribution leans and how heavy its tails are.
Arb Digital built this page as the one-stop version of a job that is usually spread across a dozen single-purpose pages. Our mean, median and mode calculator and standard deviation calculator each answer one question well, and they remain the right pages when that is genuinely all you need. This one exists for the first five minutes of working with an unfamiliar dataset, when you do not yet know which number is going to matter and pasting the data three times is wasted effort.
What This Descriptive Statistics Calculator Does
It parses your data, sorts it, and computes every summary statistic that appears in a standard introductory syllabus, plus the two that usually get skipped. Centre is covered by the mean, median and mode; spread by the range, variance, standard deviation and interquartile range; precision by the standard error; and shape by skewness and excess kurtosis. It also prints the quartiles under all three common conventions so you can reconcile an answer against whatever software produced the number you are checking.
The sample-versus-population switch changes the divisor in the variance from n − 1 to n. That single choice is the most frequent source of a small unexplained discrepancy between two tools, so it is exposed here rather than assumed. Everything else is convention-free arithmetic.
How to Use It
- Paste your numbers. Any separator works — commas, spaces, tabs, or one value per line straight out of a spreadsheet column.
- Choose sample or population. Sample is correct for almost all real data, because you are nearly always working with a subset.
- Pick a quartile method if you are checking against another tool. Exclusive matches most textbooks; linear interpolation matches spreadsheets.
- Read the headline mean, then the median beside it. The gap between them is the fastest skew check available.
- Scan the full table. The complete summary sits below the grid, including every statistic that did not fit into the four headline slots.
The Formulas and How They're Calculated
The mean is the sum divided by the count. The median is the middle value after sorting, or the average of the two middle values when the count is even. The sample variance is the sum of squared deviations from the mean divided by n − 1, and the standard deviation is its square root. The standard error of the mean is the standard deviation divided by the square root of n.
Work through the default dataset of twelve values. They sum to 192, so the mean is exactly 16. Sorted, the two middle values are 15 and 16, giving a median of 15.5. The value 15 appears twice and everything else once, so the mode is 15. The range is 22 − 11 = 11. The squared deviations sum to 122, and dividing by 11 gives a sample variance of 11.0909 and a standard deviation of 3.3303. The standard error is 3.3303 ÷ √12 = 0.9614. Under the exclusive method Q1 is 13.5 and Q3 is 18.5, so the interquartile range is 5.
Skewness and kurtosis are built from the third and fourth moments about the mean. This calculator reports the adjusted Fisher-Pearson coefficient for skewness and the excess kurtosis with the standard sample correction, which are the definitions spreadsheets use. On the default data those come out at 0.266 and −0.664: very mildly right-leaning, with tails slightly lighter than a normal distribution. The NIST/SEMATECH e-Handbook section on measures of skewness and kurtosis gives both formulas in full.
Mean, Median, Mode: Which One Is the Right Answer
All three claim to describe the centre and they answer different questions. The mean is the balance point and it uses every value, which makes it efficient and makes it fragile — one extreme observation drags it. The median is the middle rank and it ignores magnitude entirely, so it barely moves when an outlier appears. The mode is the most common value and it is the only one of the three that works on categories.
The classic demonstration is income. Take ten salaries clustered around 40,000 and add one at 900,000: the median hardly shifts while the mean jumps by tens of thousands. Neither number is wrong, but only one of them describes a typical person. The practical rule is to report the median whenever the mean and median differ by more than a small fraction of the standard deviation, and to report both whenever a reader might assume one and get the other. Our weighted average calculator handles the related case where observations carry different weights and a plain mean would understate the heavier ones.
Reading Skewness and Kurtosis Without Overreading Them
Skewness measures asymmetry. A positive value means a longer right tail — a few large values pulling the mean above the median, which is the signature of most revenue, wait-time and file-size data. A negative value means the long tail points left, which shows up in exam scores near a ceiling and in time-to-completion data with a hard deadline. Values roughly between −0.5 and 0.5 are usually treated as near-symmetric; beyond about ±1 the asymmetry is pronounced enough to affect any method that assumes normality.
Kurtosis is the more misunderstood of the two. It is often described as "peakedness", which is misleading; it is really about tail weight. Positive excess kurtosis means more extreme values than a normal distribution would produce, and that is a statement about the risk of rare large deviations rather than the shape of the middle. Both coefficients are also very unstable in small samples — with fewer than about thirty observations, a single point can swing either figure substantially, which is why this calculator prints the count prominently next to them. Treat them as a prompt to plot the data, not as a verdict, and confirm the picture with our box plot calculator or a binned view from the frequency distribution calculator.
Standard Deviation and Standard Error Are Not the Same Number
This is the most common mix-up in reported statistics and it changes the meaning of an error bar completely. The standard deviation describes how spread out the individual observations are. The standard error describes how precisely the sample mean estimates the population mean. They differ by a factor of the square root of n, so with 100 observations the standard error is a tenth of the standard deviation.
The consequence is that error bars drawn with the standard error always look tighter, and they get tighter as you collect more data even though the underlying variability of the population has not changed at all. If the question is "how variable are these units?", the standard deviation is the answer and it should not shrink with sample size. If the question is "how confident am I in this average?", the standard error is the answer. Any chart with error bars that does not say which one it used is unreadable, and this calculator gives both so the choice is explicit. To turn the standard error into an interval, pair it with our confidence interval calculator.
Why the Quartiles Differ Between Tools
There is no single agreed definition of a sample quartile. The exclusive method splits the sorted data at the median and takes the median of each half, leaving the central value out when the count is odd. The inclusive method keeps that central value in both halves. Linear interpolation abandons the halves entirely and interpolates to a fractional rank. On the same twelve numbers the exclusive method gives Q1 = 13.5 and Q3 = 18.5, while linear interpolation gives 13.75 and 18.25 — an IQR of 5 against 4.5.
That difference is small until it decides something. Because the outlier fences are built from the IQR, a point sitting near the boundary can be flagged under one convention and not the other. The rule is simply to state the method whenever a quartile is published and never to compare quartiles across tools without checking. Penn State's STAT 200 elementary statistics course notes teach the split-at-the-median approach, which is why student answers so often disagree with spreadsheet output. Our quartile calculator covers the conventions in more depth.
What Descriptive Statistics Cannot Tell You
A summary is a compression, and compression loses things. The most dangerous loss is multimodality. Two distinct groups mixed into one dataset — two customer segments, two production shifts, two measuring instruments — produce a perfectly ordinary mean, median and standard deviation while the actual distribution has two separate peaks and almost nothing in the middle. Every number on this page can look healthy while the "average" describes a value that essentially never occurs.
The second loss is order. Descriptive statistics treat the data as an unordered bag, so a series that is trending steadily upward and a series that is randomly scattered around the same mean produce identical summaries. If your observations arrived in time order, the summary is a starting point and not a description. The e-Handbook's measures of location page makes the same point from the other direction: choosing a summary statistic is a modelling decision about what you believe the data looks like. Plot before you summarise, and if there is a relationship between two variables, our correlation coefficient calculator is the next step.
Arb Digital's analysts summarise campaign and site data the way this page does — shape first, average second, conclusions last.
Browse All Free Tools Talk To Our TeamCommon Mistakes to Avoid
- Using the population divisor on sample data — dividing by n instead of n − 1 understates the variance, and the error is largest exactly where samples are smallest.
- Reporting the mean of skewed data alone — with a long tail the mean describes no one, and the median belongs beside it.
- Confusing standard deviation with standard error — they answer different questions and differ by a factor of the square root of the sample size.
- Reading skewness or kurtosis from a tiny sample — both coefficients are unstable below roughly thirty observations and can swing on a single point.
- Summarising two mixed populations as one — a bimodal dataset produces a healthy-looking summary describing a value that rarely occurs.
Related Free Tools From Arb Digital
Isolate a single measure of centre with the mean, median and mode calculator, measure spread on its own with the variance calculator, get the five-number summary alone with the five-number summary calculator, place one observation against the rest with the percentile calculator, or standardise a value with the z-score calculator. The free online tools hub lists every statistics calculator we publish.
Frequently Asked Questions
They are numbers that summarise the features of a dataset — its centre, its spread and its shape — without attempting to generalise beyond the data in front of you.
Choose sample unless you have measured every member of the group you care about. Almost all real datasets are samples, and the sample formula divides by n minus one to correct for that.
Standard deviation describes how spread out the individual observations are. Standard error describes how precisely the sample mean estimates the true mean, and it shrinks as the sample grows.
Positive skewness means a longer right tail with the mean pulled above the median. Negative means the long tail is on the left. Values near zero indicate an approximately symmetric distribution.
It compares tail weight against a normal distribution, which is set at zero. Positive values mean more extreme observations than normal would produce; negative values mean fewer.
Because there is no single agreed definition of a sample quartile. This tool prints all three common conventions so you can match whichever one the other software used.
Yes. Any real numbers are accepted, including negatives and decimals, and non-numeric entries in the pasted text are ignored rather than treated as zero.
Centre and spread are usable from around a dozen observations. Skewness and kurtosis need considerably more — thirty is a reasonable floor before either is worth interpreting.
This page explains a statistical calculation for educational purposes only. It is not medical, legal, or financial advice, and no result from it should be used as the sole basis for a decision.
