Title: Comparability and Reliability Considerations of Adequate Yearly Progress

establishing and implementing an accountability system up to the individual states, with the requirement for the system to be valid and reliable (NCLB, 2002). Determination of how well schools and districts are doing is based largely on the classification of student performance on standardized assessments. However, making classification decisions about students has been part and parcel of research on testing long before NCLB was implemented. Some early work on error in such classification systems considered the problem with a loss function framework (Traub & Rowley, 1980). Other work included focused studies of misclassification using alternate test forms (Livingston & Lewis, 1995) or the application of generalizability theory (Brennan & Kane, 1977). Large sample sizes were demonstrated to produce more accurate and precise classification (Yen, 1997). The reliability of a classification of student performance necessarily impacts AYP determinations for groups, be it subgroups within schools, schools themselves, or districts. In implementing NCLB, individual states considered the error associated with an aggregated decision and the importance of group size. In general, state-specific treatment of classification error proceeded along one of two routes. About 80 % of states use conventional confidenc