Stratified and Cluster Sampling

Plain random sampling can get unlucky. If a school is 67% freshmen and 33% seniors, a simple random sample could by chance grab mostly freshmen and badly underrepresent seniors. Splitting the population into groups first, then sampling carefully within that structure, protects against that bad luck and can make the sample cheaper or easier to collect.

Stratified sampling divides the population into strata, groups that share some trait you care about, like grade level. Then you take a separate random sample from every stratum, so each group is guaranteed representation. Cluster sampling divides the population into clusters, groups that already exist naturally, like homerooms or city blocks, and then randomly selects some whole clusters, surveying every single member inside the ones chosen. Strata are built to be different from each other but similar inside; clusters are meant to each look like a mini version of the whole population.

Both methods trade the pure randomness of a simple random sample for convenience or precision, which is why they show up alongside simple random sampling and systematic sampling as the main ways to actually collect data before you start describing or graphing it. The choice of method affects how much you can trust the sample to represent the population, which matters every time you later generalize results back to that population.