uniTEST has been developed by the Australian Council for Educational Research to assist universities with the often difficult and time consuming processes of student selection. The test has been developed to assess generic reasoning and thinking skills that underpin studies at higher education and that are needed for students to be successful at this level. The test consists of three components of 30 multiple-choice items each– Quantitative Reasoning, Critical Reasoning and Verbal-Plausible Reasoning.

In 2015 about 2000 Danish students did uniTEST in the paper-based mode, and about 1500 Danish students did the same test in the online mode. In the online mode, an English version was also delivered. The focus of this study is to explore effects the two test delivery modes and translation on item difficulty in the test.

Item response theory (IRT) Rasch model together with two most popular item estimate methods, Joint Maximum Likelihood and Marginal Maximum Likelihood, were implemented to examine the variation in item difficulties by the test delivery modes and by the two test languages.

Initial results showed that data from both test delivery modes fitted well to the Rasch model. Item difficulty patterns from both delivery modes functioned similarly to each other. Only few items in each domain showed differentially functioning by the two modes. Additionally, it seems that test language or translation showed more effects on item difficulty than the test delivery mode. However the variation in Quantitative Reasoning items was very small, where there may have low verbal loading and culture factors. Detailed discussion will be provided for individual items with large differentially functioning by test delivery modes or by test languages, in relation to item content or formats and their position in the test.