Evaluating Equating Methods for Varying Levels of Form Difference.

Area of Science:

• Psychometrics
• Educational Measurement
• Statistical Modeling

Background:

• Equating adjusts for test form difficulty differences to ensure comparable scores.
• Current equating practices often overlook the magnitude of form difficulty variations.
• Lack of guidance exists for selecting equating methods based on form difficulty differences.

Purpose of the Study:

• To investigate the impact of varying form difficulty differences on equating accuracy.
• To compare the performance of different equating methods under distinct difficulty scenarios.
• To provide evidence-based recommendations for equating method selection.

Main Methods:

• Simulation study evaluating six equating methods.
• Two common equating designs were examined: random group (RG) and common-item nonequivalent group (CINEG).
• Conditions included varying levels of form difficulty difference, from none to large.

Main Results:

• Under the RG design, mean equating excelled with no/small differences; equipercentile was superior for medium/large differences.
• For CINEG design, Tucker Linear performed best with small/medium differences; chained equipercentile or frequency estimation was optimal for large differences.
• Equating method performance is significantly influenced by the magnitude of form difficulty disparity.

Conclusions:

• The study offers crucial guidance for selecting appropriate equating methods based on form difficulty.
• Results inform testing companies on optimal equating strategies for similar and dissimilar test forms.
• Accurate score comparability relies on matching equating methods to the degree of form difficulty difference.