IRT-Skalierung der Tests im PISA-Längsschnitt 2012/2013: Auswirkungen von Testkontexteffekten auf die Zuwachsschätzung

This paper deals with the item response theory scaling of the tests in PISA longitudinal assessment 2012/2013. It presents analyses testing longitudinal measurement invariance of achievement tests in the domains of mathematics, science and reading comprehension. The analyses showed that conventional approaches of investigating longitudinal measurement invariance revealed no indication of meaningful violations of time-invariant item difficulties. On the other hand, taking into account complex booklet effects, extended analyses indicated the existence of test context effects due to the effect of item positioning. At both measurement points the PISA tests were affected by position effects; these effects being especially noteworthy and increasing over the time points in the group of nonacademic track schools. The findings of the presented study were used to derive corrections in order to counteract distortions of the results by the selective sample dropout as well as the increase in position effects. The limitations of these corrections are discussed.