Average Calculator

Mean vs. median vs. mode — which one should you use?

The mean (what most people call "the average") adds every value and divides by how many there are: for 2, 4, 6, 8, 10 that's (2+4+6+8+10) ÷ 5 = 6. It uses every number, which also makes it sensitive to outliers — one huge or tiny value can drag it far from where most of the data actually sits.

The median is the middle value once the data is sorted: for 1, 2, 3, 4 (an even count), it's the average of the two middle values, (2+3) ÷ 2 = 2.5. Because it ignores the actual size of extreme values, the median is the better "typical value" for skewed data — home prices or salaries, for example, where a few very high figures would pull the mean upward.

The mode is simply the value that shows up most often — useful for categorical or repeated data (the most common shoe size sold, the most frequent test score) rather than continuous measurements. A data set can have one mode, several tied modes (multimodal), or no mode at all if every value is unique.

Rule of thumb: use the mean for evenly-spread numeric data, the median when outliers or skew are a concern, and the mode when you care about the single most frequent value rather than a typical size.

Weighted average: when not every value counts the same

A plain mean treats every number equally. A weighted average lets some values count for more than others — the standard example is a course grade where an exam is worth more than a homework set: sum(value × weight) ÷ sum(weight). If an exam scored 90 is worth 30% of the grade and homework scored 80 is worth 70%, the weighted average is (90×0.3 + 80×0.7) ÷ (0.3+0.7) = (27+56) ÷ 1.0 = 83 — not the plain average of 90 and 80, which would be 85.

The weights don't need to add up to 1 or 100%: the division by the sum of the weights normalizes the result automatically, whatever scale you use. A weight of zero on an item still counts toward the total weight sum but contributes nothing to the numerator, so it's effectively excluded without needing to delete the row. If every weight is zero, the result is undefined (division by zero) — the calculator flags this explicitly instead of showing 0 or an error-free blank.

Geometric mean: for growth rates, not sums

The geometric mean multiplies all the values together and takes the n-th root, instead of adding and dividing: for 2, 4, 6, 8, 10 that's (2×4×6×8×10)^(1/5) ≈ 5.21. It's the right tool for rates that compound — average annual investment return, population growth rate, or any series where values multiply rather than add.

Geometric mean is only defined for strictly positive numbers. A single zero collapses the whole product to zero, and negative values make the root undefined or ambiguous — so a data set containing any value ≤ 0 shows an explicit "undefined" flag rather than a misleading number.

Population vs. sample standard deviation

Standard deviation measures how spread out the values are around the mean. There are two conventions, and mixing them up is one of the most common statistics mistakes: population standard deviation divides the sum of squared differences by n (the full count), while sample standard deviation divides by n − 1 (Bessel's correction), which corrects for the fact that a sample tends to understate the true spread of the population it's drawn from.

Use population SD when your numbers ARE the entire group you care about (every student in a class, every transaction in a closed period). Use sample SD when your numbers are a sample meant to estimate a larger, unmeasured population (a survey, a subset of measurements). With only one value entered, population SD is 0 (no spread) but sample SD is undefined, since dividing by n − 1 = 0 is impossible.

A note on averaging averages

It's tempting to average several already-computed averages together, but that only gives the correct overall average if every group being averaged has the same number of underlying values. If group sizes differ, averaging the averages silently over- or under-weights the smaller and larger groups — this is a version of Simpson's paradox. The safe approach is to go back to the raw totals: sum every original value, divide by the total count, rather than averaging the group means directly.