Table of Contents  |  Search Technical Documentation  |  References

NAEP Sample Design → NAEP 2004 Long-Term Trend Sample Design → Sampling of Schools and Students → Long-Term Trend School Stratification → Long-Term Trend Public School Stratification

NAEP Technical DocumentationLong-Term Trend Public School Stratification

The primary stratifiers for public schools were NAEP region for certainty PSU schools and PSU stratum for noncertainty PSU schools. Within these strata, type of location was the next strata categorizer:

• 1 = central city large city,
• 2 = central city mid-sized city,
• 3 = urban fringe large city,
• 4 = urban fringe mid-sized city, and
• 8 = rural area, metropolitan PSU¹.

In the certainty PSU stratum, if a school-type stratum had an aggregate measure of size corresponding to less than 2 sampled schools (i.e., the expected sample size from the stratum is less than 2), it was collapsed with the neighboring type-of-location stratum. The bullets below present the collapsing that was done for the certainty PSU stratum, for each NAEP region stratum:

• NAEP Region 1 (Northeast): Type of location 3 and 4 collapsed.
• NAEP Region 2 (Southeast): Type of location 1, 2, and 3 collapsed.
• NAEP Region 3 (Central): Type of location 1 and 2 collapsed.
• NAEP Region 4 (West): Type of location 1, 2, and 3 collapsed.

No collapsing of type of location was done for the noncertainty PSUs (within each noncertainty PSU, the number of levels of type of location was fairly low: one or two levels tended to predominate, precluding any need for a collapsing scheme).

Within the type-of-location strata, race/ethnicity percentages served as the next implicit stratifier (defined as the summation of percent blacks, percents Hispanics, and percent American Indians). For noncertainty PSUs, this was the final level of implicit stratification: race/ethnicity percentage was the final sort variable in the serpentine sort that determined the frame order. Also for certainty PSUs, if the type-of-location stratum had an aggregate measure of size less than 4 sampled schools (i.e., the expected sample size from the stratum is less than 4.0), the last stage in the serpentine sort ordering was an ordering by race/ethnicity percentage.

For certainty PSU type of location strata with a larger aggregate measure of size (4.0 or greater), race/ethnicity strata were generated, with roughly one race/ethnicity stratum per aggregate measure of size 2.0 (i.e., in terms of sampled schools). For example, a type-of-location stratum with aggregate measure of size between 4.0 up to 6.0 would have two race/ethnicity strata, a type-of-location stratum with aggregate measure of size between 6.0 to 8.0 would have three race/ethnicity strata, etc. Grade enrollment was used as a final sort within each race/ethnicity stratum, completing the serpentine sort. The idea was to use both race/ethnicity percentage and grade enrollment as implicit strata if there were enough schools within the type-of-location stratum, and otherwise use only race/ethnicity percentage within type-of-location. Of the 13 NAEP region-type of location strata defined for the certainty PSUs, 4 had aggregate measures of size less than 4.0 and had race/ethnicity percentage as the final sort variables, and 9 had aggregate measures of size greater than 4.0 and had race/ethnicity strata with grade enrollment percentage as the final sort variable.

¹ Note that type of location levels 5, 6, and 7 are not present for certainty PSUs (they are not in metropolitan statistical areas).

Printer-friendly Version