Sampling, Methodological Issues in

Probability Sampling

Probability and purposive sampling can be linked to specific worldviews and systems of beliefs that underlie how they are carried out in research, and that guide the philosophical assumptions associated with their discipline (a paradigm). Everything from how reality is perceived to how data should be collected is prescribed by the paradigm a researcher embraces during a study. Post-positivism typically encompasses probability sampling, based on the ontological belief that there is a single objective reality that everyone universally experiences. Post-positivists believe that the information gained from conducting a study to observe a phenomenon experienced by one sample can be translated or generalized to the larger population, as a whole. The sampling methods, experiments, and hypotheses are not context-specific, but can be replicated across different studies with little adaptation. Thus, the quantified results, themes, and analyses work well to explain trends across large data sets and sample sizes because there is so much information that can be boiled down to statistics.

There are five major types of probability sampling used in social and behavioral research: (1) simple random, (2) systematic, (3) stratified, (4) cluster, and (5) multistage sampling. Simple random sampling gives each item in the population an equal chance of being selected for inclusion in the sample. Simple random sampling with [Page 1532]replacement proposes that once an item is chosen it is returned to the population where it is possible to choose the item again. However, sampling without replacement is the method most often used, where once an item is chosen it is removed from the population. Systematic sampling involves the researcher setting up a sample interval (e.g., choosing every 5th person) and starting point (e.g., item 4), and using them as parameters for collecting the sample (i.e., 4, 9, 14, 19, 24, 29, 34, 39, 44, 49). Stratified random sampling attempts to generate a more representative sample of the population by assuring that certain sub-samples are included in the overall main sample. For instance, a researcher looking to sample college students may group the subsamples by year in school (i.e., freshman, sophomore, etc.) to ensure that each classification is properly represented in the sample. Cluster sampling is similar to stratified sampling in dealing with subgroups of the population; however, cluster sampling uses naturally occurring groups rather than researcher-specified, and the whole group is measured as a single item rather than the individual representing one unit. In sampling college graduation rates in the United States, for example, it could prove beneficial to take a cluster sampling of each state’s graduation rate, and possibly even further classification to state-funded or private institution. Common cluster samples can be broken down as cities, zip codes, communities, sex, and political affiliation, for example. Many studies utilizing cluster samples do so as part of multistage sampling, which involves the use of more than one probability sampling method in the same sample. After choosing initial clusters (e.g., states), it may prove beneficial for a researcher to further investigate any phenomena occurring inside the cluster, so a random sample of the cluster sample is taken. All of these techniques rely on a level of random selection or probability in choosing a representative sample, but there are fruitful ways of sampling without utilizing probabilistic methods.

...