Incorporation of content balancing requirements in stratification designs for computerized adaptive testing

Chi Keung Leung, Hua Hua Chang, Kit Tai Hau

Research output: Contribution to journalArticle

Abstract

In computerized adaptive testing, the multistage a-stratified design advocates a new philosophy on pool management and item selection in which, contradictory to common practice, less discriminating items are used first. The method is effective in reducing item-overlap rate and enhancing pool utilization. This stratification method has been extended in different ways to deal with the practical issues of content constraints and the positive correlation between item difficulty and discrimination. Nevertheless, these modified designs on their own do not automatically satisfy content requirements. In this study, three stratification designs were examined in conjunction with three well developed content balancing methods. The performance of each of these nine combinational methods was evaluated in terms of their item security, measurement efficiency, and pool utilization. Results showed substantial differences in item-overlap rate and pool utilization among different methods. An optimal combination of stratification design and content balancing method is recommended.

Original languageEnglish (US)
Pages (from-to)257-270
Number of pages14
JournalEducational and Psychological Measurement
Volume63
Issue number2
DOIs
StatePublished - Apr 2003

    Fingerprint

Keywords

  • Adaptive testing
  • Content balancing
  • Item selection
  • Stratification design

ASJC Scopus subject areas

  • Mathematics (miscellaneous)
  • Psychology(all)
  • Developmental and Educational Psychology
  • Psychology (miscellaneous)

Cite this