Choosing a formula
Population vs Sample: Which Formula When
Errors in statistics rarely announce themselves; they wait until a lecturer, a reviewer or a report QC catches them. The most common silent culprit looks harmless: two formulas, differing by a single character in a denominator. Population standard deviation divides the squared deviations by n; sample standard deviation divides them by n − 1. Pick the wrong one and everything downstream — confidence, significance, a grade — quietly shifts.
The deciding fact is not the difficulty of the formula but the ambition of the question. If your data covers every member of the group you care about, it is a population, and you divide by n. The entire class's test scores, every machine's daily output, the full census-room of readings you actually took — that is the population reality. Use it when what you have is the whole story.
When only a sample exists
Far more often your data is just a hand we pulled to guess about a larger hand we cannot see — 40 customers inferred from ten, a week of readings standing in for a year. That is a sample, an estimate of a population you did not measure. And estimates need correction. A sample is accidentally neater than the truth: it clusters its points around a mean it built itself, so it under-reports the true spread. Dividing by n − 1 — Bessel's correction — widens the answer just enough to stop that bias.
The test is 'set or sample'
Force the question into one sentence before touching a calculator: "Is this the whole set, or a part standing in for a set?" If it is literally all of it, population. If it is meant to stand in for something bigger — the population at large, the next month, the full production line — sample. That single word does the choosing for you. A standard deviation calculator makes the switch invisible: toggle the mode and re-read the same table with the corrected denominator.
The mystery crumbles fast. Investigate your own dataset once, compute it both ways, and watch the two numbers land within a whisper on 30 points but drift apart as n shrinks. The gap is the tax a sample pays for being a sample — and knowing when to pay it is the whole skill.