Table of Contents  |  Search Technical Documentation  |  References

NAEP Sample Design → Sample Design for the 2001 Assessment → National Main Assessment Sample Design in 2001 → Sampling of Primary Sampling Units (PSUs) for the 2001 National Main Assessment → Definition of PSU Strata for the 2001 National Main Assessment

NAEP Technical DocumentationDefinition of PSU Strata for the 2001 National Main Assessment

			Number of Noncertainty Strata in Each Major Stratum

Eight major strata were formed by crossing region and metropolitan statistical area (MSA) status. The primary sampling units (PSUs) were classified into four regions, each containing about one-fourth of the U.S. population. These regions, as listed below, were defined primarily by state.

• Northeast: Connecticut, Delaware, District of Columbia, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont, and Virginia

• Southeast: Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia

• Central: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin

• West: Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oklahoma, Oregon, Texas, Utah, Washington, and Wyoming

Note that those counties and independent cities in Virginia that are part of the Washington, DC-MD-VA MSA are included in the Northeast region. The remainder of Virginia is included in the Southeast region.

The 22 largest PSUs were included with certainty. The inclusion of these PSUs in the sample with certainty provided an approximately optimum, cost-efficient sample of schools and students when samples were drawn within them at the required national sampling rate. The 22 largest PSUs are presented below by region:

• Northeast: Baltimore, MD MSA; Boston-Lawrence-Salem-Lowell-Brockton, MA NECMA; New York-Northern New Jersey-Long Island, NY-NJ CMSA (excluding the part in CT); Philadelphia-Wilmington-Trenton, PA-NJ-DE-MD CMSA; Pittsburgh-Beaver Valley, PA CMSA; and Washington, DC-MD-VA MSA

• Southeast: Atlanta, GA MSA; Miami-Fort Lauderdale, FL CMSA; and Tampa-St. Petersburg-Clearwater, FL MSA

• Central: Chicago-Gary-Lake County, IL-IN-WI CMSA; Cleveland-Akron, OH CMSA; Detroit-Ann Arbor, MI CMSA; Minneapolis-St. Paul, MN-WI MSA; and St. Louis, MO-IL MSA

• West: Dallas-Fort Worth, TX CMSA; Denver-Boulder, CO CMSA; Houston-Galveston-Brazoria, TX CMSA; Los Angeles-Anaheim-Riverside, CA CMSA; Phoenix, AZ MSA; San Diego, CA MSA; San Francisco-Oakland-San Jose, CA CMSA; and Seattle-Tacoma, WA CMSA

The remaining smaller PSUs were not guaranteed to be selected with certainty for the sample. These were grouped into a number of noncertainty strata, and one PSU was selected from each stratum. In each region, noncertainty PSUs were classified as MSA (metropolitan) or non-MSA (non-metropolitan).

Within each major stratum, further stratification was achieved by ordering the noncertainty PSUs according to several additional socioeconomic characteristics, yielding 72 strata. The strata were defined so that the aggregate of the measures of size of the PSUs in a stratum was approximately equal for each stratum. The size measure used was the population from the 1990 Census.

The characteristics, available for all PSUs, that were used to define strata were as follows:

• the percent minority population,

• the percentage change in total population since 1980,

• the per capita income,

• the percent of persons age 25 or over with college degrees,

• the percent of persons age 25 or over who completed high school, and

• the civilian unemployment rate.

Up to four of these characteristics were used in any one major stratum. For each major stratum, the characteristics used were chosen by modeling NAEP PSU-level mean reading proficiency scores for the years 1988–1992 (e.g., for a particular major stratum, the characteristics might be per capita income and percent minority population, for another, civilian unemployment rate, percent of persons with college degrees, and percent change in total population, depending on what registered as significant in regression models with reading proficiency in that particular major stratum). The characteristics chosen were the best predictors of PSU-level mean reading proficiency scores in these models. An adjustment was made to re-scale 1988 data to match data from years 1989, 1990, and 1992. PSU-level mean reading scores were not needed for all PSUs in stratum. The point was to model the scores from characteristics that were available for all PSUs, so that those characteristics could be used to form efficient strata for gathering proficiency scores.

Printer-friendly Version