Modeling complex longitudinal data from heterogeneous samples using longitudinal latent profile analysis Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 22, 2019
  • Cole, Veronica Tess
    • Affiliation: College of Arts and Sciences, Department of Psychology and Neuroscience
  • Traditional approaches for examining longitudinal data tend to assume that (1) each individual's shape of change over time follows a pre-specified functional form, and (2) this same basic shape applies to all individuals under study. However, these assumptions are not always tenable in current psychological research. The current report introduces longitudinal latent profile analysis (LLPA), a mixture model which relaxes these assumptions by allowing flexible representation of both inter-individual difference and intra-individual change over time. The LLPA framework allows for this flexibility by modeling the shape of change as a function of two parameters - a time-invariant level, and a vector of time-specific deviations from that overall level - and allowing these paramters to vary categorically between individuals according to latent classes. LLPA and an extension including random effects within-class are applied to an empirical dataset concerning the development of depression in early adulthood, and results are compared to a number of traditional models. The sensitivity of LLPA to random noise in the data is then explored through a brief proof-of-concept simulation. Potential opportunities brought to bear by LLPA, as well as limitations of this approach, are discussed.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Bauer, Daniel
  • Master of Arts
Graduation year
  • 2014

This work has no parents.