Temporal and Sequence Methods

Most of what we know about learning comes from cross-sectional snapshots, yet the questions that matter most — how engagement evolves, whether strategies transfer across courses, why some students recover from setbacks while others do not — require longitudinal and temporal methods. My research tackles these questions using very large datasets that span entire degree programs, revealing patterns invisible to single-course studies.

Through sequence analysis, I have mapped the longitudinal trajectories of online learning strategies across 10 successive courses, identifying stable “intense” trajectories associated with deep learning alongside fluctuating and surface-learning pathways. Using multi-channel sequence analysis, I have shown how engagement and achievement co-evolve across multiple dimensions simultaneously. Hidden Markov models and mixture models have allowed me to uncover the heterogeneous transition dynamics of engagement states and to identify the instructional variables that trigger transitions between engaged and disengaged states. The VaSSTra method (Variables to States to Trajectories), which I co-developed, provides a general-purpose framework for converting any set of longitudinal measurements into states, sequences, and clustered trajectories — applicable at any time scale from days to years.

Selected Publications

← Back to Interests Overview