Design of Experiments > Survey Sampling
Contents:
What is Survey Sampling?
Survey sampling is selecting members from a target population to be in a sample for a sample survey. Usually the survey is some type of questionnaire (i.e. in-person, phone or internet survey).
The science of survey sampling has greatly transformed the way we view society and issues facing society. In the late 19th century the only acceptable method of gaining knowledge about a population was through a census, which surveys every member of a population. Just over a hundred years later, and the development of statistics has led to cheaper and faster ways–like survey sampling–to gauge public opinion and measure social factors.
The three parts of survey sampling are:
- Sample selection.
- Data Collection: collecting the data through mail, phone, or in-person.
- Estimation: using estimators from the collected data to make inferences about the population as a whole.
Survey Sampling: Sample Selection
Sample selection for survey samples fall into two main types:
- Probability-based samples, which chooses members based on a known probability. This uses random selection methods like simple random sampling or systematic sampling. For a list of probability-based sampling methods, see this article: Probability Sampling.
- Non-probability samples, where the probability of choosing a member cannot be calculated. Instead, non-random selection methods use the researcher’s judgment, proximity of subjects, or other non-random factor. For a list of non-probability sampling methods, see this article: Non-Probability Sampling.
Usually, probability-based samples have preference over non-probability samples. Most larger organizations and institutions publish guidelines for the types of survey sampling permitted. For example, this Whitehouse document titled “Lists of Standards for Statistical Surveys states that:
Agencies must develop a survey design…selecting samples using generally accepted statistical methods (e.g., probabilistic methods that can provide estimates of sampling error). You should justify any use of nonprobability sampling methods (e.g., cut-off or model-based samples) and you should also measure estimation error.
The “estimation error” usually refers to being able to calculate a confidence interval and margin of error.
The Kish Grid is one way to select survey participants. It assigns equal probabilities to each person and so avoids selection bias.
What is Survey Research?
Survey research is a broad term that involves collecting sample data from peoples’ responses to questions. Questions can range from a short, two-question feedback form to an in-depth personal interview about a specific topic. Three techniques commonly used:
- Questionnaires: written questions (on paper or on a computer), which can include open ended questions, closed ended questions, multiple choice questions, or questions that require the respondent to rate something on a scale (i.e. 1 to 10).
- Interviews: oral questions (in person or via a phone or computer) — closed-ended or open-ended, multiple choice or on a scale.
- Surveys: brief interviews about a specific topic. Like questionnaires and interviews, these can also come in a variety of question formats.
Advantages of Survey Research:
- It is an effective tool for collecting data from a wide variety of people about a wide variety of topics.
- It’s versatile and can provide a deeper understanding of just about any issue.
- It can be both time and cost efficient. Many questions are asked without increasing cost or time factors. If well-designed, survey research can also be relatively low cost.
- It is sometimes the only available tool for obtaining data from a large population.
Disadvantages of Survey Research
- Response rates can be very low, especially for mail surveys.
- Misleading or difficult to understand questions can taint results. Language barriers can also cause issues with responses.
- Personal interviews can be expensive, especially if the sample is spread across a wide geographic area. Cost can also be a factor with the postage involved in sending out mass-mail surveys. Telephone surveys can also quickly become cost prohibitive when factoring in the cost of training staff and payroll.
Surveys have been in popular use for decades. The odds are, if you want to find out something about a population, someone has performed survey research for that topic before you. Building on someone else’s survey research design can help you avoid many of the pitfalls of survey research, like irrelevant questions or omitting questions that are crucial to understanding the topic.
References
Brick, J. The Future of Survey Sampling. Public Opinion Quarterly. Available online here.
Schaeffer, R. (2011). Elementary Survey Sampling 7th Edition. Cengage Learning.