Sampling
Sampling
Bias: A statistical sampling or testing error caused by systematically favoring some outcomes over others.
Error: An act, assertion, or belief that unintentionally deviates from what is correct, right, or true; a mistake.
In an unbiased sample, every member of the population has an equal chance of being included in the sample. Three big sources of bias are failing to identify all members of a population (doing a study on Georgetown students without getting a list of Continuing Education students), samples of convenience (e.g. giving a survey only to your friends), and volunteerism (why is this person volunteering? The reason may influence the study).
Just because a sample is not biased does not mean it necessarily represents the population. There’s always a chance that random errors occur, called “sampling errors.” To a point, the larger the sample size, the smaller the error. However, in every sample, a point is reached in which additional participants have no significant impact on the results. This is called the point of diminishing returns. For example, an additional 10 participants in a 30-person sample should help decrease sampling errors. An additional 10 participants in a 1500-person sample does not make much of a difference.
Little variability in a population means that a small sample will still yield accurate results because there isn’t much difference or variability in what the sample has to represent. The more variability in a population, the larger the sample should be to reduce sampling errors.
Random sampling
The most basic form of sampling is known as simple random sampling, in which participants are selected based on a completely random basis, such as a random number table. Any error in simple random sampling is a result of random chance, in which the researcher might randomly get a sample that is not representative of the population. This is known as a sampling error.
Systematic sampling
The participants are arranged in some sort of order (such as alphabetically), and every Nth person is selected.
Stratified random sampling
The population is divided into different specific groups that can then be studied. This insures that each group is proportionally represented. For example, a study that is interested in racial and ethnic differences might stratify the population by race, then sample randomly within each racial group of interest to ensure each is represented adequately in the final sample.
Cluster sampling
The process for this works like simple random sampling, but instead of individuals being drawn, entire groups (clusters) are chosen at random. This can increase response and participation; however, there is a risk that the clusters are homogenous within themselves such that each cluster individually does not represent the population as a whole. For example, a study examining political opinions of Methodists may randomly sample different churches around the country and survey the members of the congregation. The risk here is that Methodists from the south may differ in many ways from Methodists on the west coast.
Purposive sampling
This is usually done for the purpose of selecting (not randomly) the individuals that will provide the best information for the study.
Snowball sampling is used for hard-to-locate participants in order to make connections.
Multi-stage sampling utilizes different methods on each level.
Sample Size:
Increasing the sample size increases precision as long as the sample is unbiased. There are four factors that influence how large a sample size should be: variability in population, results of pilot studies that indicate sample variability, costs, and occurrence of trait in population.
Back to Research and Design
References:
Patten, Mildred L. (2002). Understanding research methods: An Overview of the essentials (3rd ed.). Los Angeles: Pyrczak Publishing.
Schutt, Russell K. (1999). Investigating the social world: the Process and practice of research (2nd ed.). Thousand Oaks: Pine Forge Press.
Solso, Robert L., Johnson, Homer H., & Beal, M. Kimberly. (1998). Experimental psychology: a case approach (6th ed.). New York: Longman.
