engleski

Comparison of Classical Test Theory and Item Response Theory in Developing Knowledge Tests for Croatian National Assessment

The aim of this paper is to compare Classical Test Theory (CCT) and Item Response Theory (IRT) and to show benefits of using IRT approach in knowledge-test development. In the international project which aimed to develop instruments for Croatian national assessment, Mathematics and Croatian language tests were pretested. The samples consisted of 1100 gymnasium students. After the pre-test, CTT and IRT parameters were calculated. In analysis of results we used TiaPlus software (Heuvelmans, 2008) for CTT and OPLM software (Verhelst, Glas & Verstralen, 1995) for IRT analysis. OPLM is a model where difficulty parameters are estimated and discrimination indices are imputed as known constants, which is an extension of the Rasch model. Two types of item statistics were compared: item difficulty parameter (b) from IRT model with CTT item difficulty p value and IRT item discrimination parameter (a) with CTT item discrimination index. Choice of the main- test material was done according to the results of psychometric analysis. Pearson correlation indicated that there were considerable similarities between the item statistics obtained through CTT and IRT. Information functions calculated in OPLM were helpful when deciding which of the items with similar CTT parameters were better. Results of CTT and IRT analysis were dominantly consistent. IRT parameters were especially useful for detecting some differences between items that had very similar CTT parameters. When developing knowledge tests it is strongly recommended to use both test theory approaches.

Classical Test Theory; Item Response Theory; national assessment; knowledge tests

nije evidentirano