Probability sampling: The process of randomly selecting research participants from a target population
In quantitative research, the sample is presumed to be representative of the target population when it shares the attributes of that population. If the sample is not representative, varying degrees of sampling errors may compromise the generalizability of research findings. Probability sampling is a common approach used by researchers to ensure that samples are indeed representative. It involves randomly selecting participants from a sampling frame (the portion of the target population that is accessible to the researchers), so that each individual in that sampling frame has an equal probability of being selected.
Probability sampling is a three-step procedure. The first step is to identify the target population — school-aged children, for example. The second step is to identify the sampling frame — perhaps all school-aged children in Community X. The third step is to randomly select the required sample of children from the sampling frame, which is often too large to constitute the study sample.
Probability sampling techniques include simple, systematic, stratified and cluster random sampling. The selection of a technique is often governed by the geographical distribution of the target population and/or by the characteristics of the population that may particularly interest the researchers. For instance, for a study at a single site on a uniform population, researchers may use simple random sampling. Excel or other statistics programs are used to randomly shuffle an electronic list of all available individuals in the sampling frame to select the required sample.
Systematic random sampling is a form of simple random sampling that is especially attractive for large sampling frames. The researchers divide the accessible population (e.g., 1,000 people) by the required sample size (e.g., 100) to determine the sampling interval (i.e., 10). They then randomly select a number between 1 and the sampling interval number (e.g., 5). Starting with participant 5, they will select every tenth participant until they have recruited 100 people for their sample.
Researchers often use stratified random sampling when they believe that specific attributes must be proportionately represented in the sample. For example, if they believe that treatment response may vary depending on the severity of the disease, they may elect to recruit patients for two proportionate simple random subsamples: one for the segment of the population with the mild form of the disease and another for those with the severe form. The two subsamples are then merged to formulate the final sample.
Cluster random sampling is especially useful when the research extends over a vast geographical area that may be costly and time consuming to cover through simple random sampling. For example, if researchers are studying primary care services in Nova Scotia, they may divide primary care providers into clusters across county lines and randomly select four or five clusters from which to sample.
Remember, probability sampling is meant to minimize sampling error associated with poorly representative samples. Readers should be concerned about research generalizability if researchers fail to properly outline their probability sampling technique.
NurseONE.ca resource on this topic
- Andrew, S., & Halcomb, E. J. (Eds.). (2009). Mixed Methods Research for Nursing and the Health Sciences.
- Crosby, R. A., DiClemente, R. J., & Salazar, L. F. (Eds.). (2006). Research Methods in Health Promotion.
- Maltby, J., Williams, G., McGarry, J., & Day, L. (2010). Research Methods for Nursing and Healthcare.