Abstract
Content balancing is an important issue in the design and implementation of computerized adaptive testing (CAT). Content-balancing techniques that have been applied in fixed content balancing, where the number of items from each content area is fixed, include constrained CAT (CCAT), the modified multinomial model (MMM), modified constrained CAT (MCCAT), and others. In this article, four methods are proposed to address the flexible content-balancing issue with the a-stratification design, named STR_C. The four methods are MMM+, an extension of MMM; MCCAT+, an extension of MCCAT; the TPM method, a two-phase content-balancing method using MMM in both phases; and the TPF method, a two-phase content-balancing method using MMM in the first phase and MCCAT in the second. Simulation results show that all of the methods work well in content balancing, and TPF performs the best in item exposure control and item pool utilization while maintaining measurement precision.
Original language | English (US) |
---|---|
Pages (from-to) | 467-482 |
Number of pages | 16 |
Journal | Applied Psychological Measurement |
Volume | 31 |
Issue number | 6 |
DOIs | |
State | Published - Nov 2007 |
Keywords
- A-stratification methods
- Exposure control
- Fixed content balancing
- Flexible content balancing
- Item selection
- Two-phase content balancing
ASJC Scopus subject areas
- Psychology(all)
- Psychology (miscellaneous)
- Social Sciences (miscellaneous)