| Abstract | The National Educational Panel Study (NEPS) aims at investigating the development of competences across the whole life span and designs tests for assessing these different competence domains. In order to evaluate the quality of the competence tests, a wide range of analyses have been performed based on item response theory (IRT). This paper describes the data on scientific literacy for starting cohort 2 in kindergarten. Next to descriptive statistics of the data, the scaling model applied to estimate competence scores, analyses performed to investigate the quality of the scale, as well as the results of these analyses are presented. The test consisted of 26 items and was administered to 2,955 children. A partial credit model was used for scaling the data. Item fit statistics, differential item functioning, Rasch-homogeneity, and the tests’ dimensionality were evaluated to ensure the quality of the test. The results show that the items, except for one, exhibited good item fit and measurement invariance across various subgroups. Moreover, the test showed a high reliability. As the correlations between the two knowlegde domains are very high in a multidimensional model, the assumption of unidimensionality seems adequate. However, a justified point of criticism is that the test lacks very difficult items. But overall, the results revealed good psychometric properties of the science test, thus supporting the estimation of a reliable scientific literacy score. This paper describes the data available in the Scientific Use File and provides ConQuest-Syntax for scaling the data. (DIPF/Orig.) |
|---|
Verfügbarkeit