Research project Optimal calibration of questions in computerized achievement tests

Achievement tests to evaluate abilities of examinees become increasingly important and are used by decision makers. Several large tests are repeated periodically, for example the Swedish Scholastic Aptitude Test, national tests in school, and international tests PISA and TIMSS.

Questions (also called items) are usually pretested for future use. In this pretesting, item calibration is done which means determination of item characteristics like difficulty. The pretested items are collated into an item bank which facilitates construction of tests with a predefined difficulty and ensures test comparability over time.

In computerized achievement tests, items can be selected specifically for each individual based on the examinee’s ability. By this, the best possible information can be obtained for item calibration. We will develop statistical methods and software and implement the methods in the context of the Swedish national test. Implementation will be done in close collaboration with the PRIM-group at Stockholm University responsible for the Swedish national test in mathematics.

This research project aims to improve item calibration which increases quality for future achievement tests.