Modeling Trajectories of Latent Classes to Understand the Academic Performance of Engineering Students

This research paper investigates the grade point average (GPAs) trajectories of engineering students and provides a step-by-step guide for those interested in adapting or using the technique. The use of GPA as a measure of academic achievement is sometimes controversial, but it remains a consistently utilized measure of student success (Scheidt, M., Senkpeil, R., Chen, J., Godwin, A., & Berger, E. IEEE Frontiers in Education Conference (FIE) (pp. 1-5). 2018). A student's GPA is often used by universities to monitor their eligibility for financial aid, filter admission to colleges and departments of engineering, and determine satisfactory progress in degree attainment. In engineering in particular, a low GPA is often seen as a signal that one is not “cutting it” in the highly competitive, rigorous culture of engineering. (Godfrey, E., & Parker, L. Journal of Engineering Education, 99(1), 5-22, 2010). For many students, that signal can suggest that they should leave the major. Thus, understanding how engineering students' GPA functions over time can provide insight into students' academic outcomes. In conjunction with additional behavioral data and psychological measures, temporal changes in GPA can also help researchers identify particular supports and barriers for student success. Several techniques can be used to examine longitudinal outcomes we also have a robust collection of methodologies for identifying latent classes or clusters within larger groups. The intersection of longitudinal and latent class approaches can add further value to research about GPA trajectories and their interpretation. For instance, growth mixture modeling is used with increasing frequency to identify and model group-based changes over time (Frankfurt, S., Frazier, P., Syed, M., & Jung, K. R. The Counseling Psychologist, 44(5), 622-660, 2016). This approach allows for the identification of multiple subpopulations, description of longitudinal change within each subpopulation, and examination of differential changes among unobserved sub-populations. This approach can provide insight about patterns of academic success, as measured by GPA changes over time among and within groups of students. Along with investigating GPA changes over time, this paper is simultaneously designed to act as a resource for engineering education researchers interested in performing group-based trajectory analyses. Specifically, this paper provides a step-by-step description of exploratory latent class trajectory modeling (LCTM). The data for this example are GPAs from 489 engineering undergraduate students, as collected from institutional records, across five time points (Fall 2018 - Fall 2020). The results of the LCTM indicate a four-group solution over time, with one group starting with high GPA and decreasing slightly, a second group starting around average and decreasing slightly, a third group that started around average and decreased sharply, and a fourth group that started below average and increased sharply. Potential issues with this solution include the possible effects of missingness and attrition, as well as issues identifying the best random effect structure using the current program (e.g., whether residual variances are allowed to vary by class, allowing for random intercepts or slopes). In addition to reviewing the analysis and providing a guide and resources for other researchers, this paper will also discuss future plans for analysis with a larger sample who also provided information about a variety of non-cognitive and affective (NCA) factors in order to identify significant predictors of engineering student success.