Treating test-item nonresponse.

Abstract

This study presents the results of an empirical investigation into the effects of nonresponse on the measurement of student ability in large scale educational achievement studies. The analyses explored the bias in student and national ability estimate distributions relating to particular psychometric treatments of nonresponse. A range of replicated analyses were undertaken using data collected from a crossnational study of reading achievement. The relative merits of these treatments are summarized in conclusion, and implications for normative and methodological research and practice are considered. It is suggested that environmental and psychological antecedents of nonresponse need to be determined and that related variables be included as essential components in item response modelling.