Strategies for controlling item exposure in computerized adaptive testing with the partial credit model

Area of Science:

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

Background:

• Randomization procedures are effective for test security in CAT.
• Controlling item exposure is crucial for maintaining test integrity and fairness.
• The partial credit model is frequently used in educational and psychological assessments.

Purpose of the Study:

• To investigate the performance of four item exposure control procedures in a CAT environment under the partial credit model.
• To compare the effectiveness of Kingsbury-Zara, modified-within-.10-logits, Sympson-Hetter, and conditional Sympson-Hetter procedures.
• To evaluate these procedures based on exposure rates, overlap rates, and pool utilization.

Main Methods:

• Simulated CAT sessions were conducted using the partial credit model.
• Four exposure control procedures were implemented: no control (baseline), Kingsbury-Zara, modified-within-.10-logits, Sympson-Hetter, and conditional Sympson-Hetter.
• The Kingsbury-Zara and modified-within-.10-logits procedures were tested with both 3 and 6 item candidates.

Main Results:

• The Kingsbury-Zara and modified-within-.10-logits procedures, particularly with 6 item candidates, demonstrated comparable performance to the conditional Sympson-Hetter procedure.
• These methods effectively controlled item exposure rates and overlap.
• Pool utilization was also favorable for the recommended procedures.

Conclusions:

• The Kingsbury-Zara and modified-within-.10-logits procedures with 6 item candidates are highly recommended for partial credit CATs.
• These procedures offer a balance of simplicity and strong performance in controlling item exposure.
• Their effectiveness supports enhanced test security and efficient item pool usage in adaptive testing.