Measurement in STEM education research: a systematic literature review of trends in the psychometric evidence of scales

Abstract

Background: The objective of this systematic review is to identify characteristics, trends, and gaps in measurement in Science, Technology, Engineering, and Mathematics (STEM) education research.

Methods: We searched across several peer-reviewed sources, including a book, similar systematic reviews, conference proceedings, one online repository, and four databases that index the major STEM education research journals. We included empirical studies that reported on psychometric development of scales developed on college/university students for the context of post-secondary STEM education in the US. We excluded studies examining scales that ask about specific content knowledge and contain less than three items. Results were synthesized using descriptive statistics.

Results: Our final sample included the total number of N = 82 scales across N = 72 studies. Participants in the sampled studies were majority female and White, most scales were developed in an unspecified STEM/science and engineering context, and the most frequently measured construct was attitudes. Internal structure validity emerged as the most prominent validity evidence, with exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) being the most common. Reliability evidence was dominated by internal consistency evidence in the form of Cronbach’s alpha, with other forms being scarcely reported, if at all.

Discussion: Limitations include only focusing on scales developed in the United States and in post-secondary contexts, limiting the scope of the systematic review. Our findings demonstrate that when developing scales for STEM education research, many types of psychometric properties, such as differential item functioning, test–retest reliability, and discriminant validity are scarcely reported. Furthermore, many scales only report internal structure validity (EFA and/or CFA) and Cronbach’s alpha, which are not enough evidence alone. We encourage researchers to look towards the full spectrum of psychometric evidence both when choosing scales to use and when developing their own. While constructs such as attitudes and disciplines such as engineering were dominant in our sample, future work can fill in the gaps by developing scales for disciplines, such as geosciences, and examine constructs, such as engagement, self-efficacy, and perceived fit.