Table of Contents  |  Search Technical Documentation  |  References

NAEP Analysis and Scaling → Drawing Inferences from NAEP Results → Guidelines to Assure the Validity of Inferences Drawn From NAEP Data

Guidelines to Assure the Validity of Inferences Drawn From NAEP Data



Limiting Comparisons Involving Extreme Percentages

A key reporting goal is to assure that inferences made from NAEP data are as accurate as possible. However, under some circumstances the observed data does not allow for the computation of reliable statistics (e.g., composite and content area scale scores, percentiles, percentages of students performing at or above the achievement levels, and percentages of students indicating particular levels of background variables) or standard errors and, subsequently, these are not reported.

There are several reporting conventions used to identify whether results are reportable or not. To ensure adequate power, results have to be based on at least 62 observations. Furthermore, results have to be based on data from at least 5 replicate strata. Also, there is a convention related to extreme percentages. When estimates of percentages are close to 0 or 100, the usual normal distribution assumptions no longer apply and standard methodology cannot be used. In this case, standard errors and significance tests are not reported.

Last updated 03 July 2008 (FW)

Printer-friendly Version