Collection of statistical data can sometimes be easy. For instance, collecting the English test scores of a class does not take much time. However, if an enormous amount of data is needed, census (a study or survey of the entire population) may not be practical in terms of time and money, and a sample survey of the target population can be done. A good sample survey has to be carefully designed to ensure the sample is representative of the population.

The key to a good sample survey is to ensure that the characteristics of the sample are infinitely close to the population. However, in reality, not every sample is close to its population. A poor sample will lead to erroneous conclusions about the population, which is called bias in statistical terms. First, you have to define your target population, and then the sampling unit that constitutes your sample. A common sampling method is to first define your sampling frame, i.e. a list from which a sample is drawn and that can be a list of telephone numbers, merchandise serial numbers, etc., and then randomly select a sample from the list. It is important to note that a sampling frame may contain inappropriate samples or non-response samples that may cause bias. For example, when you do a survey among students of a university on their time spent going to the university from home and the student register is taken as the sampling frame, students who reside in dormitories would be inappropriate samples. If these students are included in the sample, it would lower the average time spent and lead to erroneous conclusions. Sometimes, it is not practical to create a sampling frame as it is costly or even impossible, and in such case, other methods should be used instead.
Got it! Go to the formula.