Survey methodology

Survey methodology is the scientific study of survey methods. It is a field of applied statistics that focuses on human-research surveys. Researchers use these methods to study how to select a sample from a larger population. They also study how to collect data using various techniques. These techniques include building questionnaires and finding ways to improve response accuracy. A survey uses specific instruments or procedures to ask questions. These questions may or may not be answered by the people being studied. The goal is to make statistical inferences about the population. This means using the small group's answers to understand the whole group.

Every single survey is made of several key parts. First, there is a sample or a full population. In a census, researchers look at the entire population rather than a small sample. Second, there is a method for collecting data, such as a questionnaire. Third, there are individual questions or items that become data. This data can then be analyzed using statistics. Surveys can focus on many different topics. They might ask about preferences, such as a favorite political candidate. They can ask about opinions, such as whether a law should change. They can also measure behavior, like smoking habits, or factual information, like income.

Success in survey research depends on how representative the sample is. The target population can be a whole country or a specific group. It might be a list of students or a professional organization. Each member of the population is called an element. If the sample does not represent the population well, selection bias occurs. This happens when some parts of the population are overrepresented or underrepresented. For example, if a population is 75% female and 25% male, a sample with 40% females is biased. To fix this, researchers use stratified random sampling. They divide the population into sub-populations called strata. Then, they draw random samples from each stratum proportionally.

There are many ways to administer a survey, which are called modes. The choice of mode depends on cost, coverage, and respondent willingness. Common modes include telephone, mail, online, and mobile surveys. There are also personal in-home surveys and mall or street intercept surveys. Sometimes researchers use mixed modes to get better results. Different methods can create mode effects. These effects change how respondents answer the questions. Choosing the right mode is a major challenge for survey methodologists. They must also train interviewers and check data files for internal consistency.

Researchers use different structures called research designs. A cross-sectional study draws a sample and studies it only once. This describes a population at one specific time. It is a predictive, correlational design that cannot show causes. A successive independent samples design draws new random samples at different times. This can show how a population changes, but not how individuals change. Finally, a longitudinal study measures the same random sample at multiple time points. This is the best way to see how individuals change over time. It can even help researchers see the effect of naturally occurring events like divorce.

Longitudinal studies are very difficult and expensive to conduct. It is hard to find people who will stay in a study for years. Many participants leave before the final assessment, which is called attrition. This attrition is often not random, which can make the sample less representative. To protect privacy, researchers might use a self-generated identification code (SGIC). These codes use items like a birth month or a pet's name. This allows researchers to link responses over time without using real names. Researchers must compare those who left to those who stayed to check for differences.

Questionnaires are the most common tools used in this field. For a questionnaire to work, it must be reliable and valid. A reliable measure produces consistent results every time it is used. One way to check this is through test-retest reliability. This involves giving the same questionnaire to a sample at two different times. A questionnaire is valid if it actually measures what it intended to measure. This includes construct validity, which measures a specific theoretical idea. Researchers also use demographic variables, like age and race, to describe their samples. They also use self-report scales to measure opinions and judgments in a way that is both reliable and valid.