Key Concepts Study Tool: Chapter 08

1. Probability and Non-Probability Samples

Probability Samples: Each unit has a known chance of being selected, and are representative, allowing for generalization from sample to population.
Non-Probability Samples: Less common in quantitative research, usually encountered with research done by polling firms or market research agencies. Not supposed to be representative of the population, and are likely to somewhat biased.

2. Random Sampling: Simple, Systematic, Stratified/Hierarchical, Cluster Samples

Simple Random Sample: To select one observation/respondent, list all possible units, number them consecutively, and then use a random number chart to select a certain number, or percentage, of units. Every unit has an equal probability of selection.
Systematic Random Sample: Units are chosen from a sampling frame and then they consistently choose every nth unit from a random starting point (e.g. every 10th observation). It is possible for others to replicate your sample as long as they have the same sampling frame.
Stratified/Hierarchical Random Sample: Series of two or more simple random samples operating within the same population. Stratify population by appropriate criteria (e.g. age). Take a simple random sample of people within each stratified group, instead of taking one simple random sample and hoping that each age group was equally represented.
Cluster Sample: It is used when researchers cannot get a complete list of the population they wish to study, but can get a complete list of groups, or “clusters,” of the population. Randomly chooses clusters instead of individuals. Usually, everyone in a cluster is included in the sample.

3. Non-Random Sampling: Convenience, Snowball, Quota Samples

Convenience Sample: Only targets individuals who possess characteristics that make them more accessible to the researcher. Also useful when pilot testing a research instrument.
Snowball Sample: Used for populations that are not easily identified, resistant to being studied, or otherwise hard to reach. The term “snowball” is used because the sample increases in size as it rolls away from its source—just like a snowball. Initial group members become informants, leading to others in their network.
Quota Sample: Non-probability counterpart to stratified samples, but unlike a stratified sample there is non-random sampling of each stratum’s units. Essentially a convenience sample, creating under-representation of less accessible groups.

4. Sampling Error

The degree of “mismatch” between a sample and the population.
Reducing sampling error will usually reduce standard error as well.
Sources of error include: non-probability sampling, inadequate sampling frame, and non-response.
Non-random sources of bias are more serious.
Probability sampling reduces sampling error and allows for inferential statistics.