Anomaly Detection Strategy for Screening Field Test Items

• In Event: 29.082 - Roundtable Session 9
  In Roundtable Session: 29.082-15 - Methodologies for Helping to Ensure Accurate Screening and Diagnostic Decisions

Abstract

We describe a screening strategy that makes use of an anomaly detection technique to identify a list of field test items for a posterior human inspection. The goal is to alleviate the workload of psychometricians and content specialists in the review process. The task of reviewing field test items is an elaborative and collaborative effort which is time consuming and labor-intensive. Moreover, this review task is often constrained by limited resources. We propose to apply an anomaly detection technique that uses the statistics elicited from the field tests to narrow down the potential flawed items before the review. With an automatic screening method, we can quickly flag items for further investigation in the posterior review phase.

Authors

• Tsung-Hsun Tsai, Research League LLC

• Yung-Chen Hsu, GED Testing Service