Measurement of multidimensional classroom composition using the concept of hypervolumes

Dokumenttyp:	Präsentation auf Konferenz
Erscheinungsjahr:	2023
Veranstaltungstitel:	EARLI 2023, 20th Biennial EARLI Conference
Veranstaltungsort:	Thessaloniki, Greece
Veranstaltungsdatum:	22.-26.08.2023
Verwandte URLs:	https://www.earli.org/assets/files/EARLI... https://www.earli.org/assets/files/EARLI...
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Fakultät für Sozialwissenschaften > Unterrichtsqualität in heterogenen Kontexten (Karst 2023-)
Fachgebiet:	370 Erziehung, Schul- und Bildungswesen
Freie Schlagwörter (Deutsch):	Heterogenität , Multidimensionale Heterogenität
Abstract:	The present study proposes a new measure of classroom composition, which allows the integration of diversity on several social dimensions into a single index. Classroom composition can thus be characterized in a more complex way, which is especially promising from an intersectional viewpoint. We adapt the hypervolume concept from ecological trait modeling, which represents students as points within a multidimensional trait space spanned by the composition variables of interest, and classrooms as so-called hypervolumes which describe the space covered by these points (Blonder et al., 2018). Hypervolumes can be characterized through their volume and average distance of students from the hypervolume centroid (dispersion). Using data from 62 fifth-grade classrooms in Germany, we investigate the relationship of hypervolume indices and single-dimension measures of classroom composition. Additionally, hypervolume indices are tested as predictors of individual student achievement in multilevel models. We find significant correlations (r = 0.20 – 0.61) between volume and dispersion of hypervolumes and range and standard deviation on single dimensions. Hypervolume indices do not significantly predict individual student achievement. Our correlative results indicate the concurrent validity of hypervolume measures. Further investigation of their predictive validity is needed and might be obtained by using larger datasets.