The sampling scheme used when the sample is large and the creation of a sampling frame taking into account all final units of the study is expensive. Before proceeding to the draw, the so-called clusters of units (most often they arise from certain customary, organizational or official classifications - for example, in the case of student examinations, these would be majors, student groups, in the case of hospital patient examinations - hospital departments, etc.). The clusters to be included in the study are then randomly drawn (in practice, the clusters are sampled with probability proportional to the size of these clusters). If all units from the randomly selected clusters are included in the final sample, the sampling scheme is called a first stage sampling. Most often, from clusters, being units of the first stage sampling, smaller clusters are drawn (with equal probabilities), even smaller ones etc., so that at the end of this chain individual research units are drawn. Depending on the number of stages, this scheme is called two-stage, three-stage, etc., or a multi-stage sampling. Therefore, it is a simple procedure that avoids the high costs of drawing up a sampling frame for the whole population of study units, but at the price of low efficiency and numerical and interpretative difficulties. It is worth adding that the cluster sampling is the more effective the more the internal diversity of the units in the clusters is bigger, and the diversity between the clusters smaller.
GUS – Departament Programowania, Koordynacji Badań i Rejestrów