Application of the EM Algorithm for Mixture Models

A developmental trajectory describes the course of behaviour over time. Iden­ tifying multiple trajectories within an overall developmental process permits a focus on subgroups of particular interest. This research introduces a SAS macro program that identifies trajectories by using the Expectation-Maximization (EM) algorithm to fit semi-parametric mixtures of logistic distributions to longitudinal binary data. For performance comparison, we consider full maximization algo­ rithms (e.g. SAS procedure PROC TRAJ) and standard EM, as well as two other EM-based algorithms for speeding up convergence. The simulation study shows that our EM methods produce more accurate parameter estimates than the full maximization methods. The EM-based methodology is illustrated with a longitudinal data set involving adolescents smoking behaviours.

Name thesis:degree_name

M Sc

Discipline thesis:degree_discipline

Epidemiology and Biostatistics

Year dc:date.issued

2010

Author dc:creator

• Chu, Man-Kee Maggie

Advisor dc:contributor.advisor

• Koval, John

• Expectation-Maximization algorithm
• Mixture models
• Binary data
• Longitudinal trajectories

Handle dc:identifier.uri

https://hdl.handle.net/20.500.14721/18744

OAI identifier oai:identifier

oai:uwo.scholaris.ca:20.500.14721/18744