Primary sampling units (PSUs) are the first stage of sampling, with the selection of schools the second stage and the selection of students the third stage. A sample of 77 PSUs was drawn from a frame of PSUs based on current Census information. After the PSU frame was created certainty PSUs (PSUs with large measures of size that make it efficient to take them with probability of selection 1) were identified and set aside.
Stratification of the noncertainty PSUs (PSUs with probabilities of selection strictly less than 1) was carried out after an analysis of NAEP 2000 operational math and science assessment data, which attempted to identify the set of PSU-level Census-based variables that had as much association with mathematics and science
assessments as possible. Measures of size and probabilities of selection were defined, and a systematic sample of PSUs was drawn.
The PSUs satisfied the following criteria:
They consisted of one county or contiguous multiple counties.
Metropolitan Statistical Areas (MSAs) were designated as PSUs even with their large size, as they were sufficiently compact in terms of their travel costs (due to higher levels of transportation infrastructure.)
PSUs did not cross state boundaries, in general.
Non-MSA PSUs in the Northeast and Southeast NAEP regions had a minimum population of 15,000 youths (age 0 to 17 inclusive) according to the 2000 Census, and in the Central and West NAEP regions had a minimum population of 10,000 youths, in general.
Other than MSAs, PSUs were to be of minimum size (defined in terms of square miles or maximum distance between points—a rough proxy for travel time) while still satisfying the minimum population constraints.