B1833
Title: Model-based clustering of longitudinal data with nominal, multidimensional outcomes
Authors: Marc Scott - New York University (United States) [presenting]
Kaushik Mohan - New York University (United States)
Jacques-Antoine Gauthier - University of Lausanne (Switzerland)
Abstract: Methods and models for longitudinal data with multidimensional nominal outcomes are somewhat limited, but they are essential to the study of the life course. In that domain, interest centres on the time and order of life course events such as having children and working full or part-time and the duration of the phases that they delineate. Typical behaviour, or typologies, are often desired, and clustering using optimal string matching algorithms or parametric models of duration in a competing risks framework are used; the appropriateness of each derives from competing goals and orientation. We focus on the latter, model-based approach to clustering dependent data such as this, positing a parsimonious data generating process (DGP) with a novel error structure. This provides us with the ability to: simulate individual trajectories; modify trajectory characteristics over time by conditioning on variables; handle multi-state trajectories and missing outcomes. Several of these goals are particularly challenging when the number of states is of moderate size and many transitions are infrequent and/or time in-homogeneous. Using the Swiss Household Panel (SHP), we demonstrate the appropriateness of a class of clustering models for sequences with heterogeneous dependence structure that provide new techniques for assessing goodness of fit as well as yield insights into social processes.
