Table of Contents  |  Search Technical Documentation  |  References

Computation of Student Nonresponse Adjustments

Student-level nonresponse was adjusted for using weighted response adjustments within a set of defined nonresponse cells. Preliminary cells based on cross-classifications of school and student characteristics were generated:

  • sample type classification, with "S2" schools not receiving accommodations as one class and "S3" schools receiving special accommodations for SD/LEP students in the other class;

  • public/private school;

  • final school nonresponse cell;

  • a dichotomous age classification, with "above-age" students in one subset and "at- or below-age" students in the other subset (for the fourth grade the cutoff birth month was October 1990; for the eighth grade, October 1986; for the twelfth grade, October 1982); and

  • a dichotomous race classification, with non-Hispanic Whites and Asian/Pacific Islanders in one subset, and Hispanics, Blacks, Native Americans, and others in the other subset.

Geography and U.S. history assessed students were put together in these cells, and these preliminary nonresponse cells were collapsed according to the following rules:

  • if the student sample size was less than 30 or the preliminary response adjustment was greater than 2.

  • if the student sample size for any replicate was less than 15.

  • Deficient preliminary cells were collapsed with "neighboring" cells in a hierarchical manner, using the reverse ordering of variables given in the list above, with deficient preliminary cells always collapsed across the race classification first, followed by age category. 

  • Cells were never collapsed over the first category of sample type, as these were treated as two separate samples.

The final set of nonresponse cells is enumerated as k=1,…,K. The number K of final nonresponse cells after collapsing are given for each grade in the table below.

The nonresponse adjustments STNRk, k=1,…,K within the final student-caused nonresponse cells were equal to the summation of wij over all eligible students, divided by the summation of wij over all cooperating eligible students:

STNR sub k equals open paren summation over i and j elements of cell k of w sub i j closed paren divided by summation over responding students i j elements of cell k of w sub i j


lowercase k equals one to uppercase K .
Distribution of nonresponse adjustments across assessed students, by grade and statistic, national main assessment: 2001
Statistic Grade 4 Grade 8 Grade 12
Sample size 15,700 22,700 21,800
Number of nonresponse cells 82 67 63
Mean 1.048 1.083 1.321
Minimum 1.000 1.000 1.023
25th percentile 1.038 1.054 1.230
Median 1.045 1.077 1.298
75th percentile 1.056 1.099 1.400
Maximum 1.299 1.356 1.844
SOURCE: U.S. Department of Education, Institute of Education Sciences, National Center for Education Statistics, National Assessment of Educational Progress (NAEP), 2001.

Last updated 17 June 2008 (MH)

Printer-friendly Version