Exposed items detection in personnel selection assessment: An exploration of new item statistic
Abstract
Detection of exposed test items has been a major concern in most testing conditions, particularly
for high-stakes testing (e.g. licensure and certification, employee selection), where failure to protect items from exposing may compromise the validity of testing outcomes. Much of the research for item exposure in computer-based tests has focused on findings ways to minimize item usage: expending the number of test items in a bank, establishing conditional item exposure controls, etc. Limited research has been carried out in generating and investigating statistics that can reveal whether items have been exposed. The purpose of this study is two-fold. First, we explore a new item statistic for detecting whether test items were known to test takers prior to testing, which was developed based on “Cumulative Sum Chart”. Second, we compare the results obtained using CUSUMP with those using the moving p-value method.
for high-stakes testing (e.g. licensure and certification, employee selection), where failure to protect items from exposing may compromise the validity of testing outcomes. Much of the research for item exposure in computer-based tests has focused on findings ways to minimize item usage: expending the number of test items in a bank, establishing conditional item exposure controls, etc. Limited research has been carried out in generating and investigating statistics that can reveal whether items have been exposed. The purpose of this study is two-fold. First, we explore a new item statistic for detecting whether test items were known to test takers prior to testing, which was developed based on “Cumulative Sum Chart”. Second, we compare the results obtained using CUSUMP with those using the moving p-value method.
