Probability Sampling Methods

In a previous post, we discussed non-probability sampling based on an article by the International Institute for Educational Planning. The three types of non-probability samples include: Judgment – Researcher chooses the sample. Convenience – “Random” sample of those nearest to researcher.

Quota – Handpicked sample based on population percentages. There are arguments to be made that non-probability samples are useful, but they often do not withstand stringent analysis – especially with regard to how samples are picked and whether or not they truly represent the sample at large. Probability sampling is much stronger.

When probability sampling is completed correctly, the sample will have no researcher introduced bias, so it has the best chance of accurately representing the population at large. Here are the three types of probability sampling outlined in “ Basic Concepts of Sample Design for Educational Survey Research .”

Simple Random Sampling Simple random sampling is the most basic form of random sampling. By design, every element within a sample has the same probability of being selected. Samples are carried out using a random algorithm, either by assigning random numbers or allowing names to be chosen at random as part of the sample.

This is the most common type of sample used in research as it eliminates any researcher introduced elements. As is the nature of all random sampling, however, there is no guarantee that the sample itself is an accurate representation of the population. Stratified Sampling Stratified sampling is a common way to address the random sample issue.

The researcher looks at population at large and breaks them off into groups, or “strata,” based on aspects of that population that may not be properly weighted in a random sample. The researcher then chooses a random sample from within that stratum in order to better represent the population while also addressing the randomization issue. Cluster Sampling Cluster sampling is a form of random sampling that is designed to decrease costs, especially in educational research.

With cluster sampling, the researchers group those with a common element (in educational research, a good example would be a classroom). They would then first randomly choose which cluster to draw their sample from, and then within those clusters select those that are going to be part of the sample.

Probability sampling – provided the sample size is large enough to draw conclusions – is the best way to ensure that your research can stand up against scrutiny. Random sampling is not without errors, but these errors are a natural part of the study design, and can be mitigated by a large enough sample.