A multitrait-multimethod study of assessment instruments for complex problem solving.

Autor/inn/en	Greiff, Samuel; Fischer, Andreas; Wüstenberg, Sascha; Sonnleitner, Philipp; Brunner, Martin; Martin, Romain
Titel	A multitrait-multimethod study of assessment instruments for complex problem solving.
Quelle	In: Intelligence, 41 (2013) 5, S. 579-596 PDF als Volltext Link als defekt meldenVerfügbarkeit
Beigaben	Literaturangaben
Zusatzinformation	Forschungsdaten, Studiendetails und Erhebungsinstrumente
Sprache	englisch
Dokumenttyp	online; gedruckt; Zeitschriftenaufsatz
ISSN	0160-2896
DOI	10.1016/j.intell.2013.07.012
Schlagwörter	Bildungsforschung; Empirische Untersuchung; Latent-Trait-Modell; Problemlösen; Logisches Denken; Komplexität; Messinstrument; Student; Deutschland
Abstract	Recently published studies on Complex Problem Solving (CPS) suggest that assessments of CPS using multiple complex systems are only moderately related to tests of classical cognitive abilities. Further, CPS assessments show incremental validity beyond tests of other cognitive abilities when predicting relevant outcomes. However, these empirical accounts have relied on single CPS assessment instruments. We do not know whether these findings will generalize to the construct level across different CPS assessment instruments. To answer this question, we tested a sample of N = 339 German university students who completed three CPS assessment instruments based on multiple complex systems (MicroDYN, the Genetics Lab, and MicroFIN) and the matrices subtest of the Intelligence Structure Test as measure of reasoning. Students further reported their school grades. Analyses including latent multitrait-multimethod models provided support for the conceptualization of CPS as a complex cognitive ability. Results indicated that different CPS assessment instruments showed sufficient convergent validity (with a consistency mostly between.50 and.60). In addition, we found evidence for the divergent validity of CPS from reasoning (reasoning predicted two CPS facets, knowledge and control, beta_KNOW = .49 and beta_CON = .53, respectively). In the prediction of academic achievement, CPS explained variance in natural science grades after we controlled for reasoning (beta_CPS = .22), whereas social science grades were not predicted. Our findings suggest that the validity of CPS generalizes across different measurement instruments. (DIPF/Orig.).
Erfasst von	DIPF \| Leibniz-Institut für Bildungsforschung und Bildungsinformation, Frankfurt am Main
Update	2015/3