skip to content

Modern Modeling Methods (M3) Conference

Bayesian Analysis using Mplus

A one-day Mplus short course on Bayesian analysis will be held on May 24 at the University of Connecticut. The host for the short course is the Educational Psychology Department, Neag School of Education. This short course will be held prior to the Modern Modeling Methods Conference, to be held at UConn on May 25-26, 2011.

Bayesian analysis has many desirable features and advantages. It performs better than maximum-likelihood analysis in smaller samples, with fewer clusters in multilevel analysis, and for models with many factors and/or random effects. Bayesian analysis also has the advantage over maximum-likelihood analysis that substantive hypotheses can be formulated in a more flexible way using informative priors as shown in:
http://www.statmodel.com/download/BSEMv4REVISED.pdf

Bayesian analysis is also an effective way to handle missing data using multiple imputation. Bayesian analysis has been incorporated into the general latent variable modeling framework of Mplus. This makes it easier than ever to carry out Bayesian analysis in practice for a wide variety of models.

The one-day course is similar to Topic 9 in the 9-topic Mplus course sequence. See:
http://www.statmodel.com/coursesequence.shtml

For videos and handouts from past courses, see:
http://www.statmodel.com/course_materials.shtml

Lecturer:  

Bengt Muthén, Professor Emeritus, UCLA

Course content

  • Advantages of Bayesian analysis over frequentist (ML, WLS) analysis
  • Introduction to Bayesian analysis using Markov chain Monte Carlo (MCMC) methods
  • Graphics for Bayesian analysis:  posterior distributions, trace plots, auto-correlation, posterior predictive checks
  • Mediational modeling using Bayesian analysis
  • Bayesian structural equation modeling
  • Examples of Bayesian analysis for high-dimensional models
  • Bayesian mixture modeling
  • Multilevel Bayesian analysis
  • Bayesian meta-analysis
  • Multiple imputation of missing data and plausible values for latent variables

For each topic, issues of model specification, identification, estimation, testing, and model modification will be discussed. Several examples will be examined. Modeling strategies will be presented. Mplus input setups will be provided and Mplus output will be used for interpretation of analysis results. The presentations will be in lecture form with no need for laptop computing.

Prerequisites

Desirable background includes topics 1-8 of the Mplus short courses (see www.statmodel.com), covering structural equation modeling, growth modeling, mixture modeling, and multilevel modeling or some experience with the MPLUS software program.  Web movies of past short courses can be watched for free.

Schedule

  • 8:00 am — Complimentary Coffee/Continental Breakfast and Registration
  • 8:30 am–10:00 am — Lecture
  • 10:00 am–10:15 am — Break
  • 10:15 am–12:00 pm — Lecture
  • 12:00 pm–1:30 pm — Lunch (participants are on their own for lunch)
  • 1:30 pm–3:00 pm — Lecture
  • 3:00 pm–3:15 pm — Break
  • 3:15 pm–5:00 pm — Lecture

Fee

The fee for the workshop is $75.00 if registering prior to April 1, 2011 and $100 if registering after April 1, 2011. Registration details and online registration will be available on this website starting January 10, 2011.


http://www.modeling.uconn.edu/index.cfm
Wednesday, May 16, 2012 - 19:13 PM