Structural Equation Modeling (SEM) with R

Institution: see Organisers & Supporters

Programme of study: International Research Workshop

Lecturer: Dr. Holger Steinmetz (University of Paderborn)

Date: Tuesday, 29/09/15 (14:30 – 18:00) – Wednesday, 30/09/15 (09:00 – 18:00)

Max. number of participants: 25

Credit Points: 5 CP for participating in the whole IRWS

Language of instruction: English

Contents:

Structural equation models (SEMs) have become a powerful tool in the behavioral sciences to test hypotheses about relationships between variables and implications of causal structures. This workshop offers an introduction to the background, principles, opportunities, and limitations of SEMs. These issues are illustrated using the lavaan package (latent variable analysis) that is run within the free software platform R. Lavaan has recently become a serious competitor to commercial software packages and is delivers almost everything a user needs to perform SEM. Participation to the course requires some basic knowledge of regression analysis, variances, covariances of variables, and inferential statistics. Knowledge of R is not necessary.

Course topics cover:

  • A short treatment of causality (the counter factual approach) and introduction to causal models and their illustration with path diagrams / causal graphs.
  • The principle behind estimating parameters and basis for evaluation the adequacy of the model (e.g., chi-square test) including Wright’s path tracing rules and Pearls d-separation.
  • Treatment and modeling of latent variables and the connection to theoretical constructs.
  • Explanation of the lavaan syntax and exercises (modeling own data / models of the participants is appreciated).
  • Reasons for misfitting models, evaluation, diagnostics, and re-specification.
  • The problem of endogeneity and the valuable role of instrumental variables in SEMs.

Required packages to be installed:

  • psych
  • car
  • Hmisc
  • MASS
  • QuantPsyc
  • Boot
  • Mnormt
  • Pbivnorm
  • quadprog
  • simsem
  • lavaan

Prerequisites for attending:

  • Basic knowledge of statistics (variance, co-variance) and regression analysis.

You have to register for the 9th International Research Workshop to participate in this course.