When finding out about a population, it isn’t always practical to survey everyone – instead, a sample is usually taken. To help make a confident predictions about the total population, samples need to be randomly selected and large enough to represent everyone.
A sample that isn’t random can lead to biased results which are unreliable when applied to the general population. Imagine taking a sample by selecting the first ten students to walk through the front gates of a school – this is an example of ‘convenience sampling’. It’s an easy way to take a sample, but it doesn’t produce a sample that represents all students (the population). Maybe there was a lot of traffic, so the first ten people all walked to school.
Another type of sampling is ‘volunteer sampling’. This type of sampling gathers information from people in the population who want to contribute data. For example, people are asked to telephone or SMS their vote for a reality TV show. This method is almost guaranteed to be biased – only interested people will phone in!
There are other types of sampling that can address the bias of convenience and volunteer sampling. Drawing names out of a hat is a common type of ‘simple random sampling’ – it gives every person in a population an equal but random chance of being part of a sample.
In any population, there are different groups called ‘strata’. When creating a sample that attempts to include people from all of these groups in the population, it is called ‘stratified sampling’. In this type of sample, each group is represented in the same proportion found in the population. When using this method to sample students from a school, the sample would have the right proportions of students from each class.
Samples give us an important insight into the qualities of a whole population, without having to examine information from every person. When the most appropriate technique is used, sampling can make powerful predictions and help us all make more informed decisions.