Statistical Modeling of the National Assessment of Educational Progress
The purpose of this book is to evaluate a new approach to the analysis and reporting of the large-scale surveys for the National Assessment of Educational Progress carried out for the National Center for Education Statistics. The need for a new approach was driven by the demands for secondary analysis of the survey data by researchers who needed analyses more detailed than those published by NCES, and the need to accelerate the processing and publication of results from the surveys.This new approach is based on a full multilevel statistical and psychometric model for students’ responses to the test items, taking into account the design of the survey, the backgrounds of the students, and the classes, schools and communities in which the students were located. The authors detail a fully integrated single model that incorporates both the survey design and the psychometric model by extending the traditional form of the psychometric model to accommodate the design structure while allowing for student, teacher, and school covariates.
The models and analysis approach are illustrated with detailed results from two NAEP surveys.The use of a full statistical model avoids the ad hoc methods that are otherwise necessary for the analysis of the data.The modeling approach is complex and computationally intensive, but less so than the existing methods used for these surveys, and it has the twin advantages of efficiency and optimality.