A set of test questions (i.e., items) grouped around a common stimulus is often referred to as a testlet. Testlets are found on the Law School Admission Test in the Analytical Reasoning and Reading Comprehension sections. This research extends current testlet theory by identifying and modeling the relationship between testlet features and test-taker response behavior on testlets. Several estimation procedures are developed, and a small simulation study shows that the procedures evaluated produce similar results.
One of the purposes of this project is to select the testlet features with the best prediction properties using text mining. This requires estimating the testlet model in the presence of a very large number of testlets and respondents. To demonstrate the feasibility of the estimation methods, a large-scale example is computed using data from 49,256 test takers who were administered a subset from 594 items nested within 100 testlets.
Request the Full Report
To request the full report, please email Linda Reustle at lreustle@LSAC.org.