Area Probability Sample

Terminology

There are several terms that are used in relation to area probability sampling that are not frequently used except in area probability and other multi-stage sampling designs. In area probability samples, the units formed for selection at the first stage are called primary sampling units (PSUs) and those for the second stage of selection are called secondary sampling units (SSUs). The units that are actually selected at these stages are called, respectively, primary and secondary selections. If there are more than three stages, the units for the third stage may be called tertiary selection units or third-stage selection units. The final unit to be selected is called the ultimate sampling unit.

PSUs, SSUs, and perhaps other units are often selected using probability proportional to size (PPS) methods. In these cases, each selection unit is assigned a measure of size (MOS). The MOS usually represents the size of the study population found in the unit. The MOS may be known or estimated or may be a function such as the square root of the population total or a composite (e.g. the sum of the total number of males plus 1.5 times the total number of females).

Reasons for Using Area Probability Designs

Many considerations can affect the choice of an area probability design for a study. One reason to use this approach could be that there is no available satisfactory list of the study population that can serve as a sampling frame. In other cases, the researchers may desire to use data about the areas as correlates in analysis of other data collected from persons or establishments. Often the choice is driven by the fact that the data being collected are best obtained (or can only be obtained) through personal contact with, or observation of, members of the population being studied. For example, (a) questionnaire items may require that the respondent be presented with visual cues as can be done in face-to-face interviewing; (b) the study requires that medical specimens be taken or anthropo-metric measurements be made; (c) the data collection involves observing behaviors, situations, or the physical environment.

If personal contact is required, cost considerations may make a clustered or multi-stage area probability sample design the most efficient, if not the only feasible design. For instance, if the survey is to collect data through personal contact with 3,000 adults in the United States, a simple random sample (or other unclustered design), even if possible, would be prohibitively expensive. An example of a more affordable design would be collecting data on 30 adults in each of 100 relatively compact areas such as metropolitan areas, counties, cities, towns, or similar administrative areas.

...