Sometimes a group is too large to ask everyone a question. Instead, you ask a smaller group and use its answers to learn about the whole group. The smaller group is a sample, and the whole group you want to understand is the population.
A useful sample gives everyone a fair chance
Suppose you want to know the favorite school lunch of all 400 students. Asking only your friends is easy, but your friends may share similar tastes. Their answers might not represent the whole school.
One better method is to put every student's name on an equal-sized slip, mix the slips thoroughly, and draw names without looking. That is a random selection. Each student has a chance to be chosen, rather than the interviewer choosing familiar people.
Random does not mean careless. You need a method that gives the population a fair opportunity to appear in the sample.
Example: estimate a school total
A random sample of 50 students includes 18 who prefer vegetable pasta. The observed fraction is To estimate the number among all 400 students, notice that 400 is eight groups of 50:
About 144 students may prefer vegetable pasta. The word “about” matters. You have not asked all 400 students, and another random sample might give a slightly different answer.
You can also calculate Multiply a sample fraction by the population size to estimate a population count.
Example: a large sample can still be unfair
To find the favorite school sport, someone asks 100 children waiting at soccer practice. That is a large group, but it is likely to contain more soccer fans than the school as a whole.
This is sampling bias: the way people are chosen tends to favor certain kinds of answers. Asking still more children at soccer practice does not fix the selection problem. A smaller random sample from the entire school can be more useful.
The question itself should also be fair. “Which sport do you enjoy most?” is more neutral than “Don't you agree soccer is the best sport?” A leading question can influence answers even when the sample was selected well.
Expect some variation
Two fair samples need not give identical results. One sample might find 18 pasta fans out of 50 and another 21. Chance can cause this difference. A sample is a tool for estimating, not a way to know every person's answer with certainty.
Try it yourself
A library wants to know which opening hours all its registered users prefer. Should it ask only people visiting on Saturday morning, or randomly select registered users from its full membership list? If 12 of 40 randomly selected users prefer an earlier opening, estimate the count among 200 registered users.
Answer: The full membership list includes people with different visiting habits. The estimate is users. It remains an estimate, not an exact count.
What to remember
Name the population, choose a sample that can represent it, and use the sample fraction to make a sensible estimate. A big sample is not automatically a fair sample.