Use of Sampling Weights in Hierarchical Models Fit to NAEP Data

This project proposes to develop tools and recommendations about appropriate methodologies for secondary users of NAEP data who wish to estimate multilevel models. The project will involve four stages: (1) the development of methodology for estimating multilevel generalized linear models from data collected using complex sample designs; (2) the fitting of multilevel linear and generalized linear models to NAEP data sets using student weights only [the incorrect method] and using the newly developed correct methods; (3) the development of graphical and diagnostic tools to help users determine if ignoring sample weights will impact a given estimation; and (4) the adaptation of the jackknife procedure currently in use for estimating sampling variance in NAEP to multilevel model estimation.