Profiles of scientific competence in TIMSS 2003: Similarities and differences between countries

The aim of the present contribution is to investigate similarities and differences of strengths in science competences between countries, based on TIMSS 2003 data. Analyses have been based on systematic investigation of patterns of p-values (percentage correct) for individual science items. Hierarchical cluster analysis has been applied to establish meaningful groups of countries. The resulting pattern of how countries cluster together into groups of increasing size, based on similarities of strengths and weaknesses are presented and discussed. As a measure of similarity between countries we have applied the Pearson correlation coefficient between the p-value residuals (i.e. each country's set of p-values, corrected for the country’s average of all items and the international item difficulty). For each of the groups of countries, average p-value residuals have been calculated to investigate characteristic features. These features are described in terms of separate measures of relative strengths according to item format, subject domain, and cognitive domain. Finally, data on relative emphases in the intended curriculum (curriculum documents) and in the implemented curriculum (percentage of topics taught) is shown to explain to a considerable degree the patterns of achievements within the different content domains