Table of Contents  |  Search Technical Documentation  |  References

NAEP Sample Design → NAEP 2003 Sample Design → 2003 National Main Assessment Sample Design → Stratification of Schools in the 2003 National Main Assessment → Stratification for Private Schools in the 2003 National Main Assessment → Stratification for the NAEP 2003 Private School Samples

NAEP Technical DocumentationStratification for the NAEP 2003 Private School Samples

The five explicit strata for the private school samples in fourth, eighth, and twelfth grades were Catholic schools (RC), Lutheran schools (LU), Conservative Christian schools (CC), other private schools with known affiliation status (OP), and private schools with incomplete affiliation status (DK). The implicit stratification within these strata was based on the nine Census divisions, urbanization status, and percent minorities (Black, Hispanic, and American Indian). Extensive collapsing of the implicit strata was carried out to assure that the expected number of schools within each implicit stratum was reasonably large.

The starting point was Census division. A division was collapsed with another division into a larger stratum within the same Census region if the expected number of schools to be sampled within the division was less than six (within the explicit stratum). The following table documents the relationship between Census division (CENSDIV) and Census region (CENSREGN) and identifies the numeric codes associated with each.

Census division (CENSDIV) and Census region (CENSREGN), by numeric code, national main assessment: 2003

CENSDIV	Division	CENSREGN	Region
1	New England	1	Northeast
2	Middle Atlantic	1	Northeast
3	East North Central	2	Midwest
4	West North Central	2	Midwest
5	South Atlantic	3	South
6	East South Central	3	South
7	West South Central	3	South
8	Mountain	4	West
9	Pacific	4	West
SOURCE: U.S. Department of Education, Institute of Education Sciences, National Center for Education Statistics, National Assessment of Educational Progress (NAEP), 2003.

A low expected number of schools could also lead to the merging of two or more regions. For example, for grade 4 Lutheran schools, the Northeast and South Census regions were collapsed into a single stratum (“1”). The two Midwest Census divisions were not collapsed (“3” and “4”). The two Western Census divisions were collapsed into their Census region (“8”).

The second stratification variable, nested within the first stratification variable, is the NCES variable “type of location.” A level of type of location is collapsed with one or more adjacent levels if the expected number of schools to be sampled within that level is less than six. The following combinations of type of location are considered permissible: 1, 2, 3, 4, 5, 6, 7, 8, 1-2, 3-4, 5-6, 7-8, 1-2-3-4, 5-6-7-8 and 3-4-5-6-7-8.

The third stratification variable, nested within the geographic and location type strata, is based on the percentage of Black, Hispanic, and American Indian students in the prior levels of strata. (Note that stratification by minority status was limited to traditionally underserved minority groups.) As above, cut points on this variable were established so that each cell had an expected number of sampled schools greater than or equal to six.

The method used to sort the data prior to selection was the serpentine sort, which executes multiple sorts within a stratum such that bordering sample units are the most similar with respect to the sort variables. This order is often used in geographic sorting. A simple example to illustrate a serpentine sort follows. Suppose that a geographic variable and type of location are the two sort variables within each stratum. Suppose further that there are three levels of the geographic variable (East, Central, and West), and three levels of type of location (small city, medium city, and large city). If, within the East geographic class, type of location is sorted from small city to large city, then the type of location sort order should be from large city to small city in the Central geographical class, and so on. The table below visually illustrates this serpentine sort example.

Serpentine sort example, national main assessment, by geographic classification and stratum: 2003

Stratum	Geographic classification	Type of location classification
1	East	Small city
		Medium city
		Large city
	Central	Large city
		Medium city
		Small city
	West	Small city
		Medium city
		Large city
2	West	Large city
		Medium city
		Small city
	Central	Small city
		Medium city
		Large city
	East	Large city
		Medium city
		Small city
SOURCE: U.S. Department of Education, Institute of Education Sciences, National Center for Education Statistics, National Assessment of Educational Progress (NAEP), 2003.

Printer-friendly Version