Sampling

A ‘sample’ is anything that is taken to be representative of who, what and/or where a researcher chooses to study. The sample used in any piece of research is one of, if not the most important aspects of any piece of research, as this will determine what data can be gathered and what conclusions (and possibly generalizations) can be drawn from this. The first stage of sampling is to select a ‘population’ – in other words, who (or what) is to be studied. For instance, researchers may decide to study football supporters. From this, they will then need to consider how they are going to select (sample) individuals and also how many.

Generally speaking, sampling can be split into two types: probability and non-probability sampling. Probability sampling (also known as random sampling) is where every individual in the population has a chance of appearing within the sample selected. Probability sampling is most useful if the researcher is undertaking ‘analytical’ or ‘scientific’ types of research, such as using a hypothesis (and a null hypothesis), as random sampling is usually helpful in ensuring that a representative sample is drawn from a population (though of course this cannot be guaranteed) and it allows for more stringent statistical testing of the data.

The easiest way to obtain a probability sample is to employ a list containing every member of the population concerned (this is called a sampling frame) and to use a computer to randomly select a certain number of individuals from this. For instance, it is may be possible to obtain a list of all of Manchester United's season ticket holders and, from this, to randomly select certain individuals to research. This is known as a simple random sample. However, it would not be possible to obtain a random/probability sample of all Manchester United supporters, as there is no list which includes all of the supporters of this team. In this instance, the researcher would probably have to rely on non-probability sampling techniques (which we will come to in a moment).

There are various forms of probability sampling, such as systematic, stratified and cluster sampling. Systematic sampling is where the researcher selects every nth person, such as every fifth or tenth, and so on, person. Stratified sampling is where the sampling frame is subdivided on the basis of a particular variable (such as age or gender) and then a sample is randomly (though not necessarily equally) selected from these (such as selecting 10 men and 20 women from the population). Cluster sampling (or multistage cluster sampling, as it is sometimes referred to) uses pre-existing ‘clusters’, such as electoral wards or districts. In this, researchers would randomly select certain ‘clusters’ and then for each of these would randomly select certain individuals to study.

Non-probability sampling is usually employed if a sampling frame does not exist from which a random sample can be drawn. Again, there are different types of non-probability sampling, such as quota, snowballing and convenience sampling. Quota sampling seeks to gather a sample based upon certain quotas. For example, the researcher may [Page 185]want to interview twice as many men as women, so they will purposefully select twice as many men in their sample. The main reason why this is usually done is to obtain some degree of representativeness in the sample. For instance, if a researcher wanted to interview ‘typical’ fans attending football matches in the UK, it would be logical to interview more men than women (as on average more men usually attend football games in the UK than women). Snowballing (or network sampling) is where one individual used in the research is asked to put the researcher in touch with others who could also be researched. For instance, one Manchester United fan might put the researcher in touch with his or her friends who are also Manchester United fans, who in turn can put the researcher in touch with friends or people they know who likewise follow Manchester United. Finally, convenience sampling is the simplest form of sampling of all and involves the researcher selecting those individuals they have access to. This is usually not representative of the wider population, but it is often the case (due to financial, time and other constraints) that we will research those we can easily get access to.

...