Abstract
Old fashioned measures of item discrimination, such as various item-test score correlations, are sensitive to many other aspects of the test data. Estimates of discrimination parameters in the most general of the binary IRT models are quite unstable. In this article, the isthmus of acceptance is proposed as a graphical tool for item analysis. It ameliorates some of the parameter stability problems by utilizing a functional representation of the item’s performance. As currently proposed, this tool is tuned to match current standards, when they are appropriately used and correctly applied. Recommendations are made as to fruitful avenues of prospective use; one use is illustrated.
Keywords
Get full access to this article
View all access options for this article.
