Home > Conventions > Program Archive > 31st APS Annual Convention > Session
APS

31st APS Annual Convention

Risks of Default Priors in Bayesian Structural Equation Models with Small Samples and Two Approaches on How to Avoid Those Risks

Friday, May 24, 2019 · Washington, DC

Oral · Methodology

We discuss risks associated with default priors in Bayesian SEM when samples are small, and demonstrate an online application to visualize this. Two solutions are proposed to overcome obstacles in the use of default Bayesian SEM: specification of informative priors obtained by expert elicitation; and a two-stage Bayesian estimation approach.

Chairs & Discussants

  • Rens van de SchootChair
    Utrecht University
  • Sarah DepaoliDiscussant
    University of California, Merced

Presentations

  1. A Non-Technical Discussion of the Impact of Default Priors in Bayesian SEM with Small SamplesSanne Smid, Sonja Winter
  2. Demonstration of an Online Educational Application on Bayesian Default PriorsSonja Winter, Sanne Smid
  3. On Elicitation of Prior Information for Latent Change AnalysisDuco Veen, Marthe Egberts, Rens van de Schoot
  4. Controlling Prior Information in Bayesian SEM: A Two-Stage ApproachHaiyan Liu, Wen Qu, Zhiyong Zhang