A Comparison of Alternative Methods for Describing Life-Course Trajectories

John R. Warren, University of Minnesota
Andrew Halpern-Manners, University of Minnesota
Liying Luo, University of Minnesota
James Raymo, University of Wisconsin-Madison

Social science researchers have increasing access to rich data that include repeated measures of individuals’ attributes at narrow age intervals over long periods of the life course. Data on age-graded “trajectories” of attributes are being used to test theories in population studies, sociology, criminology, public health, and beyond. Despite the richness of modern “trajectory” data, there is no consensus about which technique to use to (1) identify the number of trajectories in a population; (2) describe the attributes of those trajectories; or (3) decide which trajectories best describe each individual’s biography. We will model simulated trajectory data on women’s employment statuses across their careers. Do commonly used methods (i.e., group-based trajectory modeling, optimal matching analysis, grade of membership analysis, growth mixture modeling, and naïve classification) yield the same results? If not, under what circumstances do particular methods produce valid results?