Aberrance Detection Powers of the BW and Person-Fit Indices

The study compared the aberrance detection powers of the BW person-fit indices with other group-based indices (SCI, MCI, NCI, and Wc&Bs) and item response theory based (IRT-based) indices (OUTFITz, INFITz, ECI2z, ECI4z, and lz). Four kinds of comparative conditions, including content category (CC), types of aberrance (AT), severity of aberrance (AS), and the ratios of aberrant persons (AP), were implemented under the tolerance of a .05 false positive rate. Results showed that group-based indices performed better than IRT-based indices. Although the BW indices and most of the other group-based indices exhibited over 90% detection rates, the BW indices exhibited the best stability across implemented conditions. On the basis of their highly stable detection power and objective cutoffs, the BW person-fit indices were recommended for use in diagnosing students' learning issues in classrooms.