Table of Contents  |  Search Technical Documentation  |  References

NAEP Sample Design → NAEP 2002 Sample Design → 2002 State Assessment Sample Design → Stratification of Schools in Public School Sampling Frame → Stratification Variables for the 2002 State Assessment → Missing Stratification Variables

NAEP Technical DocumentationMissing Stratification Variables

The sampling frame for the NAEP 2002 state assessment was the most recent version of the Common Core of Data (CCD) file available. Schools with missing stratification variables had their data imputed.

Schools with missing estimated grade enrollment had their estimated grade enrollment set to 20¹. Once this was done, the first two stratification variables (small or large district status, and school class size) were never missing. In addition, urbanization classification was not missing for any schools in jurisdictions in which urbanization stratification was performed.

Schools with missing or questionable values—those in which the summation of the ethnicity percentages did not fall in the range 0.97 through 1.03, indicating a gross error—in minority enrollment data were assigned the average minority enrollment within their school district, five-digit ZIP Code, or three-digit ZIP Code prefix. (The mean was only imputed at the five-digit ZIP Code level if all schools were missing ethnicity percentages at the district level, and only went to the three-digit ZIP Code level if the five-digit ZIP Code mean was missing as well.) 

Schools with missing achievement data in jurisdictions and grades for which achievement data were used in stratification were assigned the median achievement data value within their urbanization and minority classifications. The achievement data were imputed only for those schools in jurisdictions and grades in which achievement data stratification was performed.

Median household income was assigned to schools in the sampling frame by merging a ZIP Code with the data in a file from Donnelly. Any schools still missing median household income were assigned the mean value of median household income for the three-digit ZIP Code prefix or county within which they were located. In some cases, imputation was not possible at the three-digit ZIP Code level, and needed to be done at the county level. There were 11 schools in Oklahoma which could not be imputed even at this level, and had median incomes imputed from the 1994 City and County Data Book, and also 255 Department of Defense Dependents Schools (overseas) and territory schools with missing median income values at the end of this process that could not be imputed and were left missing.

¹ This is an assumed lower bound for grade enrollment figures (representing one small classroom: the median classroom size is greater than 20).

Printer-friendly Version