Sampling: How do I Choose a Sample? (Part 2)
The most important consideration for choosing a sample is to be sure everyone in the population you are interested in has the same chance of being chosen. The next consideration is how you will go about gathering the sample you need. There are several ways, and there will likely be more than one way that will suit your research situation well.
The most basic of sampling methods is known as true random sampling, or sampling with replacement. This requires that every person in the population of interest has a “non-zero equal” chance of being selected. This ensures that the sample will represent the population, and is accomplished by selecting a person from the population, recording their information, and placing them back into the population before making another selection. Of course, this makes it possible that someone in the sample will be chosen again later, and we normally don’t want this to happen. To get around this, we have a slightly altered method called sampling without replacement. In this method, every sample, rather than person, has the same chance of being chosen. For example, if we need a sample of 50 people from a population of 500, every possible combination of 50 people will have the same chance of being picked as every other combination of 50 people. Many good methods of sampling are based on this basic model.
Typical methods for choosing a sample include systematic sampling, stratified sampling, cluster sampling, and there is even one called snowball sampling. With systematic sampling, some predictable system is used. For example, choosing every third (fifth, tenth, etc.) person from the population is an effective way of systematic sampling, provided you have a list. With stratified sampling, you choose the same proportion within each strata (sub-section) of your population. For example, if you want to stratify on gender from a population that is 70% female and 30% male, you simply randomly select 70% of your sample from the females and 30% from the males in the population. The example of the dentist in the previous blog was using cluster sampling. Each month chosen contains a “cluster” of patients, and those clusters became his sample.
Another common method of sampling is convenience sampling, where the researcher uses whoever is available and agrees to participate. This method, though commonly used in academia, violates the requirements of random sampling. This is the very type of thing you would see discussed in the “limitations” section of a research paper. Sometimes the researcher has no other way to get a sample, so in the limitations section she would simply explain that the sampling method used might jeopardize the generalizability of the research results. Whatever method used to select a sample, the main thing to remember is to choose it in such a way as to maximize its representation of the population you are trying to study.