When doing research on a topic of interest and PR surveys, researchers simply cannot collect data from every single individual. Instead, they choose a smaller sample of individuals that represent the larger group. If the sample is truly representative of the population in question, researchers can then take their results and generalize them to the larger group.
In statistics, a sample is a subset of a population that is used to represent the entire group as a whole. A PR survey sample will be representative if it's within the same percentage spread as the population it's trying to reproduce.
There are different ways of sampling (the process of selecting your set/subsets of population for your research); the most used ones are:
Probability sampling means that every individual in a population stands an equal chance of being selected. Because probability sampling involves random selection, it assures that different subsets of the population have an equal chance of being represented in the sample. This makes probability samples more representative, and researchers are able to better generalize their results to the group as a whole.
Within this group we identify 3 subgroups:
- simple random sample - using a software it selects random individuals
- stratified random sample - separating the sample in subgroups (by gender, age, race) and collecting random participants from each - this ensures greater accuracy as it means each subset is represented
- cluster sampling - involves creating clusters (generally, geographic clusters such as people from specific cities) that are then used for simple or stratified random sampling
Non-probability sampling, on the other hand, involves selecting participants using methods that do not give every individual in a population an equal chance of being chosen. It is more expensive and challenging - as it requires availability of respondents that match specific criteria.
Here too we can identify 3 subgroups:
- convenience sample - using respondents that are available and convenient (or cost effective)
- purposive sample - seeking out respondents that meet specific criteria (eg: men aged 18-35 that play football)
- quota sampling - one of the most used methods in online research - limiting number of respondents per groups/attributes by use of quotas
Because sampling naturally cannot include every single individual in a population, errors can occur. Differences between what is present in a population and what is present in a sample are known as sampling errors.
While it is impossible to know exactly how great the difference between the population and sample may be, researchers are able to statistically estimate the size of the sampling errors. In polls, for example, you might often hear of the margin of errors expressed by certain confidence levels.
In general, the larger the sample size the smaller the level of error. This is simply because as the sample becomes closer to reaching the size of the total population, the more likely it is to accurately capture all of the characteristics of the population. The only way to completely eliminate sampling error is to collect data from the entire population, which is often simply too cost-prohibitive and time-consuming. Sampling errors can be minimized, however, by using randomized probability testing, a large sample size or by weighting the data.
Use of filters
Filters are used in market research to identify specific groups or subgroups within a larger, more general sample. Filters can be set up using existing questions asked inside the PR survey.
As such, we could use an age filter to identify only those aged 18-35 years old – from a general population sample of 18-90 years old.
Filters allow for data segmentation – which in return – allows establishing differences in behavior/responses between an overall sample.
While samples have huge impacts on the costs and the feasibility of a research project (more niche/purposive sample will result in higher costs), filters don’t.
It is important however that filters are discussed at the beginning of the research – to allow researchers to sample correctly (eg: don’t ask for filtering on gender if only female sample was used).
When building PR surveys, it’s important to carefully consider samples and filters to genuinely represent your consumer demographic.
We can help you evaluate your next PR survey research project!
As Senior Research Analyst, Tim focuses on the quantitative services of the company, overseeing successful delivery of all projects, clients varying from Microsoft, Skyscanner and Waitrose. His areas of speciality include panel management, online communities (both creating and fostering), survey, conjoint, max-diff, data analysis, SPSS, moderation, online focus groups, semiotics and ethnographic. Before joining Atomik Research, Tim worked as Community Manager and Researcher for companies including Cision, Allegra Strategies and Channel 4.