Cluster sample

from Wikipedia, the free encyclopedia

A cluster sample (often referred to as a cluster sample) is a form of limited random selection . The population is broken down into subpopulations, the so-called clusters or clusters . With regard to the feature to be examined, the clumps should be as similar as possible (only reduced) to the population. For a cluster sample one proceeds in such a way that only some of the clusters are selected at random. If a total survey takes place in the selected clusters, one speaks of a “single-stage” cluster sample. If, however, a random partial survey is carried out in the selected clusters, one speaks of a “two-stage” cluster sample.

Often the clusters are geographically motivated. If, for example, a Germany-wide student survey is pending, you can possibly limit yourself to a full survey in a few randomly selected schools. The reduction of the survey costs is a main motive for the use of cluster samples. If the clusters are roughly the same size and the characteristic expectation values ​​in the clusters are almost identical, the characteristic value of interest can even be estimated with a cluster sample with a smaller variance than with purely random selection of the same sample size in the (large) population. However, this cannot be assumed in every case: clusters can be systematically different from one another with regard to the characteristic of interest (e.g. schools in different residential areas).

The Haßloch test market is cited as a prominent example of a cluster sample : In the small town of Haßloch in Rhineland-Palatinate, the market research company GfK has been regularly conducting market tests of new products for its customers since 1988, as well as testing commercials and print advertising, the success of which has been achieved through the nationwide introduction of these products and marketing measures. Here, however, only one single one (= the town of Haßloch) is selected from all “Klumpen” (= German localities), which is atypical for cluster samples.

See also

Individual evidence

  1. ↑ a b Göran Kauermann, Helmut Küchenhoff: Sampling: methods and practice with R . Springer, Berlin, Heidenberg 2011, ISBN 978-3-642-12317-7 , 5.2 Cluster sample, p. 160 ff . ( [1] ).
  2. ↑ Hanne Schaffer: Empirical social research for social work: An introduction . Lambertus, Freiburg im Breisgau 2014, ISBN 978-3-7841-2417-9 , 5.3.4.2 Multi-stage samples, p. 191 ( [2] ).
  3. ↑ Hartung, J. (2009). Statistics (15th edition). Oldenbourg Munich
  4. ↑ Stenger, H. (1971). Sampling theory . Physica publishing house
  5. ↑ a b Hans-Bernd Brosius, Alexander Haas, Friederike Koschel: Methods of empirical communication research: An introduction . Springer VS, Wiesbaden 2016, ISBN 978-3-531-19995-5 , 4.7.3 cluster samples, p. 71 f . ( [3] ).
  6. ↑ Mike Syzmanski: The Hassloch Experiment. The German test market. Süddeutsche Zeitung , May 19, 2010, accessed on November 5, 2019 .