NotesFAQContact Us
Search Tips
ERIC Number: ED484785
Record Type: Non-Journal
Publication Date: 2004-Dec
Pages: 51
Abstractor: Author
IRT Scale Linking Methods for Mixed-Format Tests. ACT Research Report Series 2004-5
Kim, Seonghoon; Lee, Won-Chan
Under item response theory (IRT), obtaining a common proficiency scale is required in many applications. Four IRT linking methods, including the mean/mean, mean/sigma, Haebara, and Stocking-Lord methods, have been developed and widely used to estimate linking coefficients (slope and intercept) for a linear transformation from one scale to another. These four methods have typically been used for dichotomous IRT models but can also be extended to polytomous IRT models. This paper further extends the four linking methods to a mixture of unidimensional IRT models for mixed-format tests. The development in the present study is intended to be as general as possible so that each linking method can be applied to mixed-format tests using any mixture of the following five IRT models: the three-parameter logistic model, the graded response model, the generalized partial credit model, the nominal response model, and the multiple-choice model. A simulation study is conducted to investigate the performance of the four linking methods extended to mixed-format tests. Overall, the Haebara and Stocking-Lord methods yield more accurate linking results than the mean/mean and mean/sigma methods. The simultaneous linking using all items with different formats is compared to the linking through items of a "dominant" item format. When the nominal response model or the multiple-choice model is used to analyze data from mixed-format tests, limitations of the mean/mean, mean/sigma, and Stocking-Lord methods are described.
Publication Type: Reports - Research; Tests/Questionnaires
Education Level: Higher Education
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: American Coll. Testing Program, Iowa City, IA.
Grant or Contract Numbers: N/A