What These Statistics Measure and Where Calculators Disagree
Mean, median and mode are the three standard ways to describe the middle of a data set, and they answer slightly different questions. The mean is the arithmetic average, the median is the value sitting in the physical centre of the sorted list, and the mode is whatever value appears most often. Range adds a crude sense of spread by subtracting the smallest value from the largest. Students meet all four in a single chapter, analysts reach for them before any modelling, and teachers use them to summarise a class. Reporting only one is how data gets misread: add a single extreme value to a list of seven and the mean can move by thousands while the median barely shifts at all.
The formulas are short. Mean x̄ = Σxᵢ ÷ n, where xᵢ is each value and n is the count. Median is the middle of the sorted list when n is odd, or the average of the two middle values when n is even. Mode is every value tied for the highest frequency — plural, because a data set can genuinely have two or more, and there is no mode at all when every value occurs once. Spread is where the real trap lies. Population variance σ² = Σ(xᵢ − x̄)² ÷ n divides by n and describes a complete group. Sample variance s² = Σ(xᵢ − x̄)² ÷ (n − 1) divides by n − 1, called Bessel's correction, and is the unbiased estimate when your numbers are a sample from something larger. Standard deviation is the square root of either. This tool prints both, side by side and labelled, rather than quietly picking one.
Take the data set 10, 2, 38, 23, 38, 23, 21, which is the worked example calculator.net publishes. The sum is 155 across 7 values, so the mean is 155 ÷ 7 = 22.1429. Sorted it reads 2, 10, 21, 23, 23, 38, 38, making the median 23. Both 23 and 38 appear twice, so this data set is bimodal and both are modes — a calculator returning just one is wrong here. The range is 38 − 2 = 36. Population variance comes to 151.2653 with σ = 12.299, while sample variance is 176.4762 with s = 13.2844. That gap of nearly one full unit between σ and s is exactly what a student loses marks over. Quartiles by the inclusive method give Q1 = 15.5, Q3 = 30.5 and IQR = 15, with Tukey fences at −7 and 53, so nothing here is an outlier. Geometric mean is 16.4128 and harmonic mean 8.8922.
Four situations make this worth bookmarking. A teacher marking a class test wants the mean for the report and the median to check whether one very low score is dragging the average down. A biology student writing up an experiment must report the sample standard deviation, not the population one, because their measurements are a sample of all possible measurements. An analyst cleaning a spreadsheet uses the 1.5 × IQR fences to isolate suspicious entries before running anything else — on 1, 2, 3, 4, 5, 6, 7, 8, 9, 100 the upper fence is 14.5, so 100 is flagged immediately. And anyone averaging growth rates or ratios needs the geometric mean, since averaging percentages arithmetically overstates compounded growth.
One genuine gotcha deserves attention: there is no single agreed definition of a quartile. Excel ships two functions, QUARTILE.INC and QUARTILE.EXC, and R offers nine methods, all defensible and all capable of returning different answers on the same numbers. On the integers 1 to 10 the inclusive method gives Q1 = 3.25 and Q3 = 7.75, while the exclusive method gives 2.75 and 8.25 — and because the IQR changes, so do the outlier fences. This tool lets you choose, names the active method on screen and shows the other method's figures beside it, so a spreadsheet mismatch is explainable rather than alarming. Fences follow Tukey's Exploratory Data Analysis (1977); mean definitions follow Wolfram MathWorld. Your data set is never uploaded.
