Representative Sampling
Statisticians collect data and arrive at conclusions for various hypotheses. In an ideal world where costs, time, and resources were endlessly available, they would undertake a survey of the entire population in which the respondents meet the criteria of the study. Since this is not feasible, they do the next best thing, they choose a sample that represents the population and collect data. The findings from the sample are then extrapolated to the population.
Representative sampling is not true random sampling in which each unit has an equal chance of being selected for the sample. Instead, each unit selected for the sampling will have the characteristics of the population under study.
It is important to document what constitutes the representative sample, that is, the criteria on which the sample is considered representative.
In the pharmaceutical industry, the representative sample for a batch of medicine vials could be a vial from the start, middle, or end of the production run.
Collecting Representative Sample
The common ways to collect a representative sample are:
- Simple random sampling
- Quota sampling
- Probability sampling
- Non-probability sampling
Learn more about sampling methods.
Simple random sampling ensures minimum selection bias. If the selection method involves judgment, then it is important to go over the details of the criteria and expertise of the selector who will pick out the samples.
The representative sample can be neither too big nor too small. It also must represent subsets within the sample adequately. If the representative sample is too large, it will mean an escalation of costs, time and other resources. The benefits of surveying a larger than required representative sample are not statistically significant.
Similarly, if the representative sample is too small, the results will not be statistically significant. The sample will not be truly representative of the population. The decisions based on such a study will be poor, thanks to the poor quality of data.
For a study with a population of 20,000, with a confidence level of 95%, a margin of error of 5%, and a population proportion of 50% the required sample size is 377. This means at least 377 or more surveys are required for the study to have a confidence level of 95% and that the real value is within ±5% of the surveyed value.
