Alternative Linear Item Response Theory Observed-Score Equating Methods.

Area of Science:

• Psychometrics
• Educational Measurement
• Statistical Modeling

Background:

• Item response theory observed-score equating (IRTOSE) is a standard technique in large-scale testing.
• Existing IRTOSE methods have limitations, prompting the need for alternative approaches.

Purpose of the Study:

• To empirically evaluate three novel linear IRTOSE methods against the traditional IRTOSE approach.
• To assess the performance of these methods using simulations and real-world data.

Main Methods:

• Comparison of traditional IRTOSE with three alternative linear IRTOSE methods.
• Empirical evaluation through Monte Carlo simulations.
• Analysis of real testing data to compare equated scores.

Main Results:

• Local linear (kernel) IRTOSE methods demonstrated minimal bias and favorable loss measures.
• Exclusive reliance on traditional linear IRTOSE resulted in substantial bias.
• Significant discrepancies in equated scores were observed across different methods when applied to real data.

Conclusions:

• Local linear IRTOSE methods offer a promising alternative to traditional approaches.
• Traditional linear IRTOSE methods are not recommended due to potential for high bias.
• The choice of IRTOSE method can substantially impact score equating outcomes.