Statistics Calculator

Descriptive statistics — intuition and pitfalls

The mean is the arithmetic average: every value carries equal weight, so a single extreme outlier can pull the mean far from where most data sit. The median sorts all values and takes the middle (or averages the two middle values when n is even); it is robust to outliers and often better for skewed incomes or house prices. The mode is the most frequent value — useful for categorical codes or discrete ratings; datasets can be multimodal or have no mode if every value appears once.

Variance measures average squared distance from the mean. For a sample drawn from a larger population, dividing by n − 1 (instead of n) gives an unbiased estimate of population variance — the classic Bessel correction. If your list is the entire population, dividing by n is appropriate. Standard deviation is the square root of variance, returning units to the same scale as your measurements, which makes it easier to interpret next to the raw numbers.

Quartiles split sorted data into quarters. IQR (Q3 − Q1) captures the spread of the middle half and underpins outlier rules like 1.5×IQR fences. Geometric mean multiplies values and takes the n-th root — common for growth rates; it requires all inputs positive. Harmonic mean is the reciprocal of the average of reciprocals — natural for rates such as “average speed” when distances match.

Histograms group values into bins; changing bin width can change the story slightly — we use equal-width bins across the observed range for a compact view. Box plots encode median, quartiles, and extremes; they excel at comparing batches side by side. When every number is identical, spread measures are zero — we surface a note so you know the tool is working, not stuck.

In many real populations, mean salary exceeds median salary because a long right tail of very high earners lifts the average while the midpoint stays closer to typical pay. That is why news articles often cite medians for “typical” experience. For percent work and scientific notation, pair this page with our Percentage and Scientific calculator.

Outliers deserve explicit thought: a single corrupted sensor reading, a typo, or a jackpot bonus can dominate the mean. Before trusting a mean, glance at min, max, and the box plot — if the whiskers are tight but one point is miles away, investigate or winsorize in your formal analysis. Sample size matters for stability: variance estimates wobble more when n is tiny; our sample variance uses n−1 precisely so small samples do not underestimate spread on average.

Reporting in papers and dashboards often shows mean ± SD; readers may wrongly assume normality. Median and IQR complement means when distributions skew. For repeated measurements on the same units (before/after), consider whether paired tests apply — this calculator stays descriptive and does not infer p-values, but good summaries are the prerequisite for any inference.