Skip to content

Research

Detecting fraud in large-scale testing. Test security has become an essential topic in the past two decades as the results of high-stakes testing are increasingly used to make critical educational and resource allocation decisions. Concurrently, the incentives to engage in testing fraud have also increased. Consequently, the companies that develop and administer such tests, school districts, and state departments of education have grown concerned about the validity (i.e., accuracy) of reported test scores. Although ideally,  test fraud should be prevented, it is difficult to know when fraud occurs (i.e., during or following test administration). Therefore, there has been an increased need for statistical methods that can be used to screen large-scale item response data to identify irregularities. Dr. Zopluoglu’s work in this area focuses on developing new methods to detect test fraud (integrating machine learning algorithms) as well as evaluating the performance of currently existing methods.

Continous Response Model. Continuous Response Model (CRM; Samejima, 1973), has not received as much attention in the IRT literature as the dichotomous and polytomous models and is somewhat underutilized in practice. We encounter continuous response outcomes more frequently than thought in certain educational and behavioral settings, and CRM is a viable, and valuable, psychometric model to make inferences about the items and people with continuous response outcomes. Dr. Zopluoglu’s work in this area focuses on technical aspects of CRM, its calibration, and practical applications.