The project spans four years and the aim for the project is to contribute to higher quality of future achievement tests. This implies better characterization of examinees’ abilities and a better possibility to compare performance over time.
Large achievement tests to evaluate abilities of examinees become more and more important, both on a national (e.g. Swedish högskoleprovet and national tests) and on an international
level (e.g. PISA, TIMSS). These tests are repeated regularly and it is usually desired to pretest questions (also called items) for intended future tests. One task in this pretesting, which is called item calibration, is to determine item characteristics like difficulty. While achievement tests previously have been conducted as paper-and-pencil tests, now there are a development towards computerized tests.
For computerized tests, new and more efficient ways to calibrate items need to be developed. The project’s objectives are to develop statistical methods for efficient item calibration in computerized achievement tests, to develop software to facilitate the methods and to implementat the methods in the context of the Swedish national test in mathematics and other achievement tests.
