Skip Navigation Bar

National Information Center on Health Services Research and Health Care Technology (NICHSR)

Assessing the Results
  Page 1 of 2
Perspectives also shape the standards researchers use when they design and execute data collections.

Does the sample have to be representative of some larger population?

Do the researchers assume that conditions are homogeneous enough to allow inferences from observations of a targeted group—as for example when reporting on a selected number of reported cases of sexually transmitted diseases?

Are the researchers collecting material using a clinical experiment or case controlled study? If so, they may substitute controls [i.e. comparative observed cases] for statistical representativeness as a way of supporting inferences to a broader population.

Probability Sampling
Statisticians widely rely on probability samples—These kinds of samples are constructed using methods that allow the samples to project the results from their small samples to the entire population for which estimates are required. They are called probability samples because they rely on a design that allows samplers to know the chance that an individual will be selected and then use this information in the projection. Probability samples also let samplers estimate the uncertainity {or sampling erroe} associated with any statistic derived from the procedure. For more information see:Kish Leslie.Survey Sampling

Statisticians widely rely on sampling procedures that allow them to report results for a larger population because they know the probability of individuals being selected and can adjust for this in their reports. The need for these types of samples critically depends on the goal of making inferences to a defined and specified population.
  Page 1 of 2