A Comparison of IRT Observed Score Kernel Equating and Several Equating Methods.

Area of Science:

• Psychometrics
• Educational Measurement
• Statistical Modeling

Background:

• Equating is crucial for comparing test scores across different test forms.
• Item response theory (IRT) offers advanced methods for test equating.
• Evaluating the performance of various IRT equating methods under different conditions is essential.

Purpose of the Study:

• To compare the accuracy and stability of IRT observed score kernel equating with other equating methods.
• To investigate the impact of sample size and test length on equating accuracy.
• To assess the influence of data simulation methods on equating results.

Main Methods:

• Item response theory (IRT) observed score kernel equating, equipercentile equating, IRT observed score equating, and kernel equating were compared.
• Pseudo tests and pseudo groups were used to ensure comparability of IRT data simulation results.
• Identity equating and large sample single group rule served as criterion equating for evaluating local and global indices.

Main Results:

• IRT observed score kernel equating demonstrated superior accuracy and stability in random equivalent groups.
• IRT observed score equating exhibited the lowest systematic and random errors in non-equivalent groups with an anchor test.
• Equating errors decreased with shorter test lengths and larger sample sizes, though the latter had a negligible effect.

Conclusions:

• IRT observed score kernel equating is recommended for random equivalent groups designs.
• IRT observed score equating is preferred for non-equivalent groups with anchor test designs.
• Further research should explore improvements in equating methods and data simulation techniques.