NotesFAQContact Us
Search Tips
ERIC Number: ED464937
Record Type: Non-Journal
Publication Date: 2002-Apr
Pages: 22
Abstractor: N/A
Reference Count: N/A
Item Calibrations for Licensure Tests with Multiple Specialty Components.
Huang, Chi-Yu; Lohss, William E.; Lin, Chuan-Ju; Shin, David
This study was conducted to compare the usefulness of three item response theory (IRT) calibration packages (BILOG, BILOG-MG, and PIC) for examinations that include common and specialty components. Because small sample sizes and different mean abilities between specialty components are the most frequent problems that licensure/certification programs may encounter, this study emphasizes these issues. The three-parameter logistic model item calibrations for 360 items from a large-scale test were assumed to be the known item parameters and were used to generate data. Results show that BILOG, BILOG-MG, and PIC are equally successful in estimating item parameters when specialty groups have the same ability level However, BILOG-MG performs better than the other two packages when dealing with specialty groups with different ability levels. The study also suggests that the precision of item parameter estimates is not seriously affected when specialty groups have different ability levels. The specialty groups sample size has more effect on the precision of estimated item parameters when it reduces to 250. Although item characteristic curve plotting for some items in this study has shown that these estimation errors may be acceptable, the generalization of the results from this study to real testing situations should be made cautiously. (SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A