Implementing the a-Stratified Method with b Blocking in Computerized Adaptive Testing with the Generalized Partial Credit Model

The purpose of the current study was to generalize the use of the a-stratified method with b blocking (Chang, Qian, & Ying, 2001; denoted STR_B) to computerized adaptive testing (CAT) with the generalized partial credit model (Muraki, 1992). Chang et al. (2001) showed that STR_B performs well in reducing highly discriminating items' exposure rates and increasing lower a items' usage while maintaining measurement precision in CAT with a dichotomous item pool. The characteristics of STR_B have not yet been investigated in a CAT containing polytomous items. In the current study, the method of selecting items based on maximum information with Sympson-Hetter exposure control (Hetter & Sympson, 1997) and the original a-stratified method (Chang & Ying, 1999) were compared to STR_B in CAT with the generalized partial credit model. Additionally, this study examined the effect of the length of CATs on the performance of each method. The results of this study showed that applying STR_B in CAT with the generalized partial credit model increased item usage in the pool and controlled item exposure rates, while maintaining measurement precision, especially for longer tests.