A Continuous a-Stratification Index for Item Exposure Control in Computerized Adaptive Testing.

Area of Science:

• Psychometrics
• Educational Measurement
• Computerized Adaptive Testing (CAT)

Background:

• Item overexposure in CAT can compromise test score validity.
• Existing stratification methods use fixed, non-overlapping groups based on item discrimination parameters.
• High item exposure rates necessitate improved control mechanisms in adaptive testing.

Purpose of the Study:

• To introduce a novel continuous stratification index for computerized adaptive testing.
• To integrate item exposure control directly into the item selection index.
• To eliminate the need for predefined, discrete strata in item banking.

Main Methods:

• Development of a continuous stratification index incorporating exposure control.
• Simulation studies to compare the new index with existing stratification methods.
• Evaluation of psychometric properties including ability estimation bias and mean squared error.

Main Results:

• The continuous stratification index effectively controls item exposure rates.
• Simulation results indicate reduced ability estimation bias compared to traditional methods.
• The new method demonstrates comparable or improved performance in controlling item exposure.

Conclusions:

• Continuous stratification offers a more flexible and effective approach to managing item exposure in CAT.
• This method enhances the validity of test scores by mitigating overexposure issues.
• The proposed index provides a viable alternative to discrete stratification for item selection.