Latent Profile Analysis

In person in Calgary | Next offered July 2024 during Rocky Mountain Methodology Academy

Latent profile analysis (LPA) is a family of statistical models that can be used to identify unobserved, heterogenous, and qualitatively distinct subgroups in one’s data. LPA, as well as other forms of mixture models, are increasingly being used in numerous areas of behavioural science to address substantively important research questions. LPA, and the mixture modeling framework, provides an intuitive and model-based approach to classify individuals, teams, organizations, etc., based on the assumption that an observed sample of data includes a mixture of subgroups characterized by distinct distributions of scores on variables. As well, modern LPA models have the flexibility to incorporate predictors and outcomes of subgroup membership, as well as more complex analytical components involving mediation, moderation, and conditional effects.

This course will introduce LPA, focusing on applications in the social, educational, health, and management sciences. This course will provide participants with the theoretical and conceptual background and applied analytical skills needed to specify an appropriate analytical model, interpret the results, and thoroughly address research questions using LPA. This course offers lectures that feature worked examples, numerous hands-on activities and practice sessions, as well as ample opportunities to discuss participants’ own LPA-based research ideas.

INSTRUCTOR: Matthew McLarnon PhD

In this course, participants will learn about the underlying principles and the practical applications of LPA and other forms of mixture models. The topics covered will include

  • Brief overview and background on the Mplus software
  • Conceptual background on LPA and other mixture models such as cluster analysis and latent class analysis (LCA)
  • Distinguishing between person-centered and variable-centered analyses
  • LPA model identification, model fitting, model selection
  • Diagnosing and troubleshooting common issues
  • Including and interpreting predictors and outcomes in an LPA model
  • Introduction to advanced methods of integrating auxiliary analytical models with mediation, moderation and/or conditional effects within an LPA model
  • Visualization of network data
  • Introduction to multi-group analyses (i.e., profile invariance) and longitudinal analyses (latent transition analysis [LTA], growth mixture modeling [GMM])

This course will be delivered in person in Calgary over two days and will meet from 9am to 5pm each day.

Although LPA will be presented as a modeling framework that can be applied across different software platforms (e.g., LatentGold, SAS [with PROC LCA], or R [packages: tidyLPA, MCLUST, polCA, or OpenMX]), this course will focus on its implementation in the Mplus software environment given Mplus’ flexibility and advanced features. Because this is a hands-on course, learners are encouraged to bring a laptop to class with a recent version of Mplus installed, but this is not a requirement. All datafiles and syntax will be available to attendees so that all material can be downloaded onto personal laptops.

This course will be helpful for researchers in any field – including psychology, sociology, education, business, human development, social work, public health, communication, and other fields that use social science methodology – who want to understand how to apply and leverage LPA and other mixture models to explore their substantive research questions. Although this course will include a primer and background on the Mplus software, participants should be comfortable conducting analyses covered by a typical introductory-level graduate-level statistics course (i.e., familiar with ANOVA and multiple regression). Previous experience with Mplus is not required though would be helpful.

Upon completing this course, participants will

  • understand the conceptual and theoretical background on LPA, and other mixture models and person-centered analyses
  • know how to make informed choices when specifying, fitting, evaluating, interpreting, and troubleshooting LPA models
  • be able to implement modeling extensions to include predictors and outcomes, and other advanced analytical components (e.g., mediation, moderation, conditional effects) in an LPA model and interpret the results
  • understand the basic ideas underpinning extensions to advanced applications of LPA models
  • be able to implement the LPA in Mplus and, to a lesser extent, other software capable of estimating some kinds of LPA models.

A certificate of completion from the Canadian Centre for Research Analysis and Methods is provided at the end of the course.