Predictors of Science Subject Discipline Identities: A Statistical Analysis

Abstract

This quantitative study (n = 247) explores whether preservice science teachers express science-specific identities that reflect multiple areas of their beliefs (e.g., purpose for science teaching, inclusion of science-technology-society-environment issues into science teaching, and nature of science) as well as other individual characteristics (e.g., focus of university training, perception of self within professional communities, and interest in becoming a teacher). Hierarchical cluster analysis showed a three-cluster solution representing three subject-specific identities: Model Citizen, Model Science Teacher, and Model Non-Science Teacher. Additional analysis (multinomial logistic regression) revealed cluster membership associated with preservice science teachers’ most comfortable teaching subject.

Résumé

Cette étude quantitative (n = 247) vise à déterminer si les futurs enseignants de sciences expriment une identité spécifique aux différentes disciplines scientifiques, reflétant des aspects multiples de leurs valeurs (par exemple la motivation qui les pousse à l’enseignement des sciences, l’inclusion de questions liées aux sciences, technologies, société et environnement dans leur enseignement, ou encore la nature des sciences) ainsi que d’autres aspects plus personnels (par exemple l’orientation de leur formation universitaire, leur image de soi au sein des communautés professionnelles et leur intérêt personnel pour l’enseignement). Une analyse par regroupements hiérarchiques met en évidence trois regroupements représentant trois identités spécifiques: le citoyen modèle, l’enseignant modèle en sciences et l’enseignant modèle dans les domaines non scientifiques. Une analyse plus approfondie (par régression logistique multinomiale) montre que les regroupements sont liés aux disciplines dans lesquelles les enseignants en formation se sentent le plus à leur aise.

Notes

1. Given the complexity of the theoretical model of identity, with as many as eight factors measured by 68 items, our sample size of 247 was not sufficient to proceed with the suggested “split-half” analysis, in which the data are randomly divided in two, with exploratory factor analysis conducted on one half of the data, which then informs the confirmatory factor analysis on the second half (Loehlin, 1998). In fact, a confirmatory factor analysis of all possible factors was not possible even with the entire set of 247 cases, because that would violate the general accepted rule of at least 5 cases per parameter estimated. For this reason, confirmatory factor analysis of separate sets of items was intended merely as a check against the exploratory analysis. Because both exploratory and confirmatory factor analyses as well as Cronbach's alpha supported inclusion of items within each index, our approach to forming the factors seemed justified.