A Comparison of Label Switching Algorithms

• In Event: Poster Session 15
  In Poster Session: Important Issues in Model Selection, Measurement Invariance, Reliability, and Mixture Models

Abstract

Simulation studies involving mixture models inevitably aggregate parameter estimates and other output across numerous replications. A primary issue that arises in these methodological investigations is label switching. The current study aims to compare several label switching corrections that are commonly used when dealing with mixture models. A growth mixture model is used in this simulation study, and the design crosses three manipulated variables—number of latent classes, latent class probabilities, and class separation, yielding a total of 12 conditions. Within each of these conditions, the accuracy of a priori identifiability constraints, a priori training of the algorithm, and two post-hoc algorithms developed by Tueller, Drotar, and Lubke (2011), and Cho (2013), respectively, are tested to determine their accuracy.

Authors

• Kristina Cassiday, University of Maryland - College Park

• Jeff R. Harring, University of Maryland - College Park