Surveying or survey research involves sampling people or other populations. A large enough group needs to be sampled in order to draw inferences on the purpose of the survey. The sample will be representative if the population it is drawn from possesses the characteristics relevant to the study, i.e., are present in the sample in the same way they are in the population. The sample does not have to be representative in all aspects but it should be representative in those characteristics most important to the study.
Sample error or bias occurs when the sample is not representative. Common biases in population studies are age, sex, socioeconomic states, or ethnicity.
Sample error: random error produced by chance, e.g., sampling males and females and many more females are randomly sampled than males, e.g., telephone sample and more females are at home and answering the telephone.
Sample or selection bias: selectively and consistently choosing one option over the other, e.g., selecting females more often than males when the study includes both sexes.
Types of sampling:
Simple random sampling--generally a random number table is used; if the sample size is small, names can be written on slips of paper and drawn from a container.
Stratified random sampling--some of the variables are already known, e.g., age, ethnicity, socioeconomic status. The researcher ensures that these variables will be included in the sample. Requires a large population to draw from but the resulting sample can be smaller than with the simple random method.
Cluster sampling: sampling that selects from groups or clusters, e.g., geographic area, cities, states, organizations, institutions, that exist in the parent population. A random sample is drawn from the cluster (a heterogeneous sample is divided into a number of smaller more homogeneous clusters). It is subject to sample bias if the clusters are not representative of the parent population.