A Latent Class Analysis Approach to the Identification of Doctoral Students at Risk of Attrition

Open Access
- Author:
- Stevens, Samantha
- Graduate Program:
- Psychology
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- February 10, 2021
- Committee Members:
- Jonathan Emdin Cook, Thesis Advisor/Co-Advisor
Jes Matsick, Committee Member
Dawn Paula Witherspoon, Committee Member
Kristin Ann Buss, Program Head/Chair - Keywords:
- latent class analysis
social identity threat
intervention
higher education
attrition - Abstract:
- To advance our understanding of doctoral student experiences and the high attrition rates among STEM doctoral students, we examined the psychological profiles of different types of doctoral students and conducted a preliminary examination of the effects of two social psychological interventions. We used latent class analysis (LCA) on self-reported psychological threat data from 1081 incoming doctoral students across three universities and found that the best-fitting model delineated four threat profiles: Engaged and Confident, Relaxed and Nonchalant, Engaged but Worried, and Disengaged and Worried. These profiles predicted outcomes measured at the beginning and end of students’ first semester of graduate school that may influence attrition risk, including differences in academic preparation (e.g., prior attainment of a master’s degree), self-evaluations (e.g., academic self-control), attitudes towards graduate school and academia (e.g., burnout), and interpersonal relations and perceived fit (e.g., sense of belonging). The demographic distributions of profiles also differed, with groups more likely to face social identity threat (e.g., women) being overrepresented in higher threat profiles (i.e., Engaged but Worried and Disengaged and Worried students) and underrepresented in lower threat profiles (i.e., Engaged and Confident and Relaxed and Nonchalant students). Moreover, students completed randomly assigned belonging intervention, affirmation intervention, or control writing exercises early in their first semester and we preliminarily examined effects of each intervention by threat profile on outcomes at the end of the first semester. We did not find robust evidence for intervention efficacy overall or difference in efficacy by class. We conclude that LCA may be useful to identify students at high risk of negative outcomes and that future work should further investigate the appropriateness of tailoring interventions by student threat profile to ultimately promote retention.