|
Abstract:
|
NAEP data can be analyzed in detail using a publicly available online tool. With such access, disclosure risk increases sharply and government standards require preventive action, preferably by data swapping. Two characteristics complicate the implementation of a naïve data swap for NAEP. First, random swaps are not only required to have no impact on several levels of margins of survey indicator distributions, but also of proficiency estimates. Second, the probabilistic nature of the sample puts certain subgroups at greater risk, even at the highest level of aggregation. The solution chosen in NAEP is a procedure that aims to protect two types of margins and incorporates probabilistic swap sample selection. Results from 2005 NAEP data show that relatively small swapping rates can provide reasonable disclosure risk prevention without severe reductions in data utility.
|
