Introduction to Structural Equation Modelling using Mplus
CCSR offers 5 free places to research staff and students within the Faculty of Humanities at the University of Manchester and the North West Doctoral Training Centre.
Course Leader: Dr Nick Shryane and Professor Tarani Chandola
Course Requirements: Participants should be familiar with statistical modelling using linear regression and binary logistic or probit regression.
Structural Equation Models (SEM) amalgamate regression analysis, path analysis and factor analysis, allowing for more richly detailed statistical models to be specified and compared to data than by using these techniques individually. Historically, SEM models were confined to the analysis of continuous observed data, limiting their usefulness in applied social research, where many phenomena are inherently discrete or are measured only with coarse-grained instruments. Advances in recent years have made SEM methods for categorical data available to applied researchers. This course aims to train quantitative social scientists to use the Mplus programme in the application of structural equation modelling techniques to non-continuous observed data. The course also aims to integrate approaches that assume latent dimensions of variation (e.g. factor analysis) with approaches that assume unobserved groups or categories (e.g. latent class analysis).
Provisional Course Syllabus
Day 1: Modelling continuous and discrete observed variables
AM: Learning to use Mplus to perform linear, binary logistic and probit regression analysis
PM: Using Mplus for Path Analysis, with a particular emphasis on mediation models.
Day 2: Modelling continuous and discrete latent variables
AM: Models with continuous latent factors: Confirmatory Factor Analysis for continuous observed variables and Item Response Theory analysis for categorical observed variables.
PM: Models with categorical latent factors: Latent Class Analysis and Finite Mixture models
Day 3: Full Structural Equation Models and their Applications
AM: Multiple Indicators, Multiple Causes (MIMIC) models; Differential Item Functioning; and Multiple-Group modelling
PM: Latent Growth Models for longitudinal data
The course emphasizes conceptual understanding rather than mathematical and statistical derivations. The course aims to highlight the opportunities, assumptions and limitations in applying SEM techniques to applied social research problems.