An Introduction to Latent Class Analysis

About

This workshop will involve a mixture of interactive lecture-type sessions and practical examples using real-life epidemiological datasets in Mplus.

Course leader

Professor Gary Adamson

Target group

No prior knowledge of latent class analysis or Mplus is expected; however, attendees should have a basic understanding of the concept of using latent variables within psychological and social science research.

Course aim

Many important concepts in the disciplines of psychology and other social sciences, for example personality, quality of life, or prejudice, cannot be directly observed (i.e. they are hidden or latent constructs). Researchers often attempt to measure these concepts using standardised questionnaires, which are assumed to be imperfect indicators of the latent construct of interest. These observed indicators are assumed to be caused by the latent variable; therefore, covariation among these observed measures is expected. The patterns of interrelationships among observed measures can be explored and analysed using latent variable modelling.

A number of latent variable models are used widely in the behavioural and social sciences – the most common of which is the factor analytic (FA) model. The main difference between the traditional FA model and the latent class (LC) model lies in the nature and distribution of the latent variable. For the FA model, the latent variable is continuous and normally distributed, whereas the LC model assumes as categorical latent variable with a multinomial distribution. Use of the LC model has mushroomed in recent years largely due to the increased tendency to collect data at either the nominal or ordinal level of measurement.

Fee info