Abstract: Item Response Theory was used to analyze the Emotional Competence Scale which consisted of 9 subtests and 132 5-point Likert-type items. Factor analysis showed the first eigenvalues extracted from each subtests were significantly larger than the second ones. All the items in the subtests showed high level model-data fit when 5-grade Item Response Models were applied. The study also showed that the item parameters had hypothesized relationship with CTT item statistics and were invariant across subject groups. All the results above support the notion that IRT is feasible to analyze Likert-type items. In the study test reliability coefficient can be estimated more correctively through IRT method.

Key words: Item Response Theory, Graded Response Model, Likert Scale, model-data fit