Lacking measurement invariance in research self-efficacy: Bug or feature?

Abstract

Psychological factors play an important role in higher education. With respect to students' understanding of scientific research methods, research self-efficacy (RSE) has been studied as a core construct. However, findings on antecedents and outcomes of RSE are oftentimes heterogeneous regarding both its theoretical and empirical structures. The present study helps disentangle these findings by (a) establishing and validating an integrated, multi-dimensional assessment of RSE and (b) introducing a developmental perspective on RSE by testing the impact of the disciplinary context and academic seniority on both mean level and latent structure of RSE. The construct validity of the new measure was supported based on RSE assessments of 554 German psychology and educational science students. Relations to convergent and discriminant measures were as expected. Measurement invariance and LSEM analyses revealed significant differences in latent model parameters between most sub-groups of training level and disciplinary context. We discuss our findings of measurement non-invariance as a feature rather than a bug by stressing a process-oriented perspective on RSE. In this regard, we conclude potential future directions of research and RSE theory development, alongside implications for methods education practice in higher education.

Coding Decoded: Exploring Course Achievement and Gender Disparities in an Online Flipped Classroom Programming Course

“…Academic self-concept (t1) was assessed using three items from three subscales of the SASK (Scales Academic Self-Concept) [78], which measures an academic self-concept relating to an external frame of reference. SASK was chosen because of its validity and reliability and the possibility of using it for specific domains [81,82]. Students assessed their study abilities and their study-related skills compared to the requirements and demands of their study subject on a 7-point Likert scale.…”

Estimating Local Structural Equation Models

“…LSEM was also applied to moderator variables different from age. Klieme and Schmidt-Borcherding ( 2023 ) employed LSEM to explore whether there is noninvariance for indicators of research self-efficacy regarding different training levels of students operationalized as the number of studied semesters. Weiss et al ( 2020 ) investigated the threshold hypothesis of creativity by handling intelligence as a continuous moderator in LSEM.…”