When a target population is sufficiently large that the entire population cannot realistically be surveyed, a sample of the population can be drawn to represent characteristics, attributes, or behaviors of the whole target population.
There are two general types of sampling methods for determining how respondents will be selected: probability sampling and non-probability sampling.
Probability sampling involves the random selection is applied to all levels of the sampling process, where every element in each sampling frame has a known, non-zero chance of being selected. Random selection makes it more likely that all individuals within a sampling frame have the same probability of being sampled (or asked to participate in a survey) and that the characteristics of the respondents are representative of characteristics of the target population. Probability sampling methods include, random sampling, systematic sampling, and stratified sampling. Examples for each of these are described briefly below.
- Random sampling: From a list of all school districts in California, a sample of 300 districts are selected based on a sorted random number assigned to each district in the list. (Random number functions are generally used for the generation of a list of random numbers and are commonly found in many software programs. The list is then sorted by the newly generated random number and the required number of sampling units are determined.)
- Systematic sampling: A sample of 30 households is selected from all households in a 10 block sampling frame by starting from a randomly chosen starting point and then selecting every nth household (or 5th for example).
- Stratified sampling: All school districts in California are divided into three strata: small, medium, and large student populations, and 100 districts are randomly selected within each strata for a total of 300 districts.
Non-probability sampling involves the selection of respondents in a non-random manner. Convenience sampling, snowball sampling, and quota sampling are non-probability sampling methods because all members of the target population do not have the same chance of being selected. The findings drawn from a non-probability sample cannot be extended to a target population because the sample may only represent the characteristics of a limited segment of the population.