A school has 1,000 students and wants to know how long they spend on homework, using a sample of 50. Asking the first 50 students through the door at 7:30 a.m. is easy, but early arrivals may be more organized, or may be there for morning sports practice. That is a convenience sample, and its answer can be off however carefully the questions are asked.
Random methods fix this in different ways. A simple random sample (SRS) numbers all 1,000 students and lets a random number generator pick 50, so every group of 50 is equally likely. A stratified sample first splits the school into grades, the strata, and picks randomly within each: if sixth graders are 30% of the school, 15 of the 50 come from sixth grade. A cluster sample randomly picks a few whole homerooms and surveys everyone in them. A systematic sample takes a random starting point on the class list and then every 20th name.
Stratified and cluster sampling sound alike but make opposite moves. Stratified takes some students from every group, which guarantees each grade appears. Cluster takes every student from a few groups, which is cheaper but works well only if each homeroom is a mix of the whole school. Neither is about treatments: deciding by coin flip which sampled students try a new study app is experimental design, not sampling.
Think of it like this. Sampling is tasting soup. Stir the pot and one spoonful tells you about the whole pot, as a random sample does. Skim only the top and you taste the oil that floated up, like a convenience sample. A spoonful from each layer is a stratified sample. Whether the pot was stirred matters far more than the size of the spoon.
Core idea. Sampling methods are the systematic ways to choose a subset of a population — SRS, stratified, cluster, systematic — versus biased convenience sampling.