Psychometric Properties of the Science Self-Efficacy Scale for STEMM Undergraduates

TL;DR

This study validates a science self-efficacy scale for undergraduates, showing it works well across diverse groups.

Contribution

The study provides a comprehensive validation of the Science Self-Efficacy Scale using item response theory.

Findings

The Science Self-Efficacy Scale is unidimensional with high reliability (ωh = 0.86, marginal reliability = 0.91).

Item parameters were consistent across race/ethnicity but varied slightly by gender identity for one item.

The scale is valid for use across diverse student populations in science fields.

Abstract

Biomedical research training initiatives need rigorous evaluation to achieve national goals of supporting a robust workforce in the biomedical sciences. Higher science self-efficacy is associated with the likelihood of pursuing a science-related research career, but we know little about the psychometric properties of this construct. In this study, we report on a comprehensive validation study of the Science Self-Efficacy Scale using a robust sample of 10,029 undergraduates enrolled across 11 higher education institutions that were part of a biomedical training initiative funded by the National Institutes of Health in the United States. We found the scale to be unidimensional with an Omega hierarchical (ωh) reliability coefficient of 0.86 and a marginal reliability of 0.91. Within the item response theory framework, we did not detect variation in item parameters across undergraduates’ race/ethnicity; however, one item had parameters that varied across gender identity. We determined that the Science Self-Efficacy Scale can be employed across undergraduates enrolled in science, and researchers can use the scale across a diverse group of students. Implications include ensuring that the scale functions consistently across diverse populations, enhancing the validity of conclusions that can be drawn from survey data analysis. Validating this construct with item response theory models strengthens its use for future research.

Figures