ERIC Number: ED308247
Record Type: Non-Journal
Publication Date: 1988-Apr
Reference Count: N/A
A Cluster-Based Method for Test Construction. Research Report 88-3.
A new test construction method based on integer linear programming is described. This method selects optimal tests in small amounts of computer time. The new method, called the Cluster-Based Method, assumes that the items in the bank have been grouped according to their item information curves so that items within a group, or cluster, are interchangeable. Introducing this assumption may reduce the number of decision variables in the model. Because of the assumption of interchangeability, the accuracy of the method will also be reduced, but this reduction is small. The process of test construction is independent of the particular item response model used. The basic model for cluster-based test construction using information functions is described. How the model can be generalized to allow for subject matter constraints is shown. Methods of dealing with additional test specifications and selecting individual test items are described. The algorithm for solving the integer linear programming models is presented. Six test construction problems were then analyzed, and the results of these analyses were presented in three tables. In each case an item bank of 1,000 items was used. These examples revealed that the basic model for cluster-based test construction works well in terms of central processing unit time and accuracy. Additional examples of the test construction method describe the construction process for a selection and a diagnostics test and the process of constructing four parallel tests. Two additional tables give the characteristics of these tests. (SLD)
Descriptors: Computer Assisted Testing, Item Banks, Latent Trait Theory, Mathematical Models, Test Construction, Test Items, Test Selection
Bibliotheek, Department of Education, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands.
Publication Type: Reports - Evaluative
Education Level: N/A
Authoring Institution: Twente Univ., Enschede (Netherlands). Dept. of Education.