In this post, I'll explain what random sampling is, the different types of random sampling you might come across and an alternative to it.

What is random sampling?

Random sampling is a way of selecting a sample of observations from a population in order to make inferences about the population. For example, exit polls from voters that aim to predict the likely results of an election. Random sampling is also known as probability sampling. The main forms of random sampling are simple random sampling, stratified sampling, cluster sampling, and multistage sampling. Samples that are not random are typically called convenience samples.

The key aspect of random sampling lays in its name. The selection of observations must occur in a 'random' way, meaning that they do not differ in any significant way from observations not sampled. It is typically assumed that statistical tests contain data that has been obtained through random sampling.

Simple random sampling

Simple random sampling is the most straightforward approach for getting a random sample. It involves picking a desired sample size and selecting observations from a population in such a way that each observation has an equal chance of selection until the desired sample size is achieved

Stratified random sampling

Stratified random sampling involves breaking a population into key subgroups and obtaining a simple random sample from each group. These subgroups  (e.g., males under 30, females under 30, males 30 or over, and females 30 or over) are called strata. For example, if you want a sample size of 200, then you can pick samples of 50 from each strata. The required sample size for each stratum will be designed either to match known population proportions, or to over represent key subgroups of interest.

The main benefit of stratified sampling over simple random sampling is making sure that you have good sample sizes in key subgroups.

Cluster sampling

Cluster sampling is like stratified random sampling, except that the population is divided into a large number of subgroups (e.g., hundreds of thousands of small subgroups); then some of these subgroups are selected at random, and simple random samples are then collected within these subgroups. These subgroups are called clusters.

Typically, the purpose of cluster sampling is to reduce the costs of data collection. This is achieved by defining clusters according to ease of access (e.g., a suburb may be a cluster if door-to-door sampling, or a household may be a cluster if phone interviewing).

Multistage sampling

Multistage sampling is where a combination of the above techniques is used. A combination of different sampling techniques (e.g., stratification with cluster sampling within each strata).

Alternatives to random sampling

Convenience sampling refers to approaches where considerations of simplicity rather than randomness determine which observations are selected in a sample (e.g., asking the opinions of people that you know, rather than a random selection from the population).