NAEP Technical DocumentationSchool Sample Selection for the 2019 State Assessment

			Computation of Measures of Size School Sample Sizes: Frame and New School Evaluation of the Samples Using State Achievement Data

The sampled schools for the fourth- and eighth-grade public school state assessments in mathematics and reading came from two frames: the primary public school sample frame constructed from the Common Core of Data (CCD) and the supplemental new-school sampling frame. Schools were sampled from each school frame with probability proportional to size (PPS) using systematic sampling. Prior to sampling, schools in each frame were sorted by the appropriate implicit stratification variables in a serpentine order. A school's measure of size was a complex function of the school's estimated grade enrollment. Schools whose measure of size was larger than the sampling interval could be selected or “hit” multiple times. Schools with multiple hits were selected with certainty and had larger student sample sizes.

For the CCD-based frame, schools were sampled at a rate that would yield specific target student sample sizes for each jurisdiction. At grades 4 and 8, all jurisdictions, except Puerto Rico, had a target student sample size of 5,700 students. The goal was to obtain 4,900 assessed students: 2,200 students for the reading operational assessments, and 2,700 students for the mathematics operational assessments. Puerto Rico had a target student sample size of 4,000 students. By design, Bureau of Indian Education (BIE) schools were not part of the state assessments this year. However, separate BIE school samples were selected based on target student sample sizes that were large enough to ensure that BIE schools were sufficiently represented in the national samples.

The schools in the new-school frame were sampled at the same rate as the CCD-based school frame.

Prior to selection, schools were deeply stratified in each jurisdiction to ensure that the school sample distribution reflected the school population distribution as closely as possible, with regard to the stratification variables, to minimize sampling error. The success of this approach was shown by comparing the proportion of minorities enrolled in schools (based on CCD values for each school), median income, and urban-centric locale (viewed as an interval variable) reported in the original frame against the school sample.

In addition, the distribution of state assessment achievement scores for the original frame can be compared with that of the school sample for those jurisdictions for which state assessment achievement data are available, as was done in the evaluation of the samples using state achievement data.

Printer-friendly Version