Sampling lets you estimate a population's characteristics without measuring every member. The quality of the estimate depends on how the sample is selected, not merely on its size.
Population, sample, and the question
The population is the entire group you want to describe. The sample is the smaller group actually observed. If a town wants the proportion of all residents who support a new park, a survey of 200 residents is a sample; the town's residents are the population.
Be precise about the population. “Students in this class” and “all students in this district” are different targets. A sample that is representative of one may not represent the other.
A simple random sample gives every possible sample of the chosen size an equal chance of selection. For school-level problems, selecting names randomly from a complete list is a typical implementation.
Example: scale a sample proportion
In a random sample of 80 students, 28 walk to school. The estimated walking proportion is For a school of 600 students, the predicted count is
The estimate scales the proportion, not the raw count. Adding the difference between 600 and 80 to 28 would have no statistical meaning. The calculation also assumes the sampled students reasonably represent the whole school.
Another random sample might produce a different fraction. This is sampling variability: sample results change by chance even with a good method.
Bias is a selection problem
A convenience sample uses people who are easy to reach. Surveying only students in the walking club will likely overestimate the school's walking proportion. A voluntary-response survey lets people decide whether to participate; those with strong views may be more likely to respond.
Undercoverage occurs when some groups are omitted from the selection process. Nonresponse occurs when selected people do not provide data. Either can distort a result if the missing people differ in relevant ways.
A larger biased sample can produce a very precise estimate of the wrong group. Increasing size reduces ordinary random variation; it does not automatically repair bias.
Example: compare two methods
A school wants opinions about a homework policy. Method A polls 300 students who voluntarily click an online link. Method B randomly selects 100 students from the enrollment list and follows up to obtain responses.
Method B has a stronger basis for representing the school, despite being smaller. Method A may attract students with unusually strong opinions. You cannot decide the better method from sample size alone.
A neutral question matters too. Asking “How many hours did you spend on homework yesterday?” collects more direct information than “Doesn't our unfair homework policy take too much time?”
Try it yourself
A random sample of 120 households includes 45 with a dog. Estimate the number with dogs among 2,000 households. Explain why surveying only visitors to a dog park would be unsuitable.
Answer: so the estimate is Dog-park visitors are more likely to own dogs, so that selection would tend to overestimate dog ownership.
What to remember
Define the target population, assess the selection method, calculate the sample proportion, and describe any projection as an estimate. Random variation and systematic bias are different limitations.