RMME/STAT Joint Colloquium

Bayesian Covariance Structure Modeling: An Overview and New Developments

Dr. Jean-Paul Fox
University of Twente

Friday, April 23rd, at 2:00PM ET

https://uconn-cmr.webex.com/uconn-cmr/j.php?MTID=m51820f42c5c0cf72fc3979c5bccd49a2

There is large family of statistical models to understand clustered or hierarchical structures in the data (e.g., multilevel models, mixed effect models, random effect models). The general modeling technique is to use a latent variable (i.e., random effect, frailty parameter) to describe the covariance among clustered observations, where the strength of the covariance is represented by the latent variable variance. This approach has several disadvantages. It is only possible to describe positive within-cluster correlation (similarity), and not dissimilarity (Nielsen et al., 2021). Sample size restriction and model complexity are often implied by the number and type of latent variables. Furthermore, the latent variable variance is restricted to be positive, which leads to boundary issues at/around zero and statistical issues in evaluating data in support of a latent variable. A new approach for modeling clustered data is Bayesian covariance structure modeling (BCSM) in which the dependence structure is directly modeled through a structured covariance matrix. BCSM have been developed for various applications and complex dependence structures (Fox et al., 2017, Klotzke and Fox, 2019a, 2019b; Mulder and Fox, 2019). This presentation gives an overview of BCSM and discusses several applications/new developments: (1) BCSM for measurement invariance testing (Fox et al., 2020); (2) BCSM for identifying negative within-cluster correlation and personalized (treatment) effects in counseling; and (3) BCSM for interval-censored, clustered, event-time data from a three-armed randomized clinical trial investigating coronary intervention. This talk discusses prior specification, the multiple-hypothesis-testing problem, and computational demands.

 

---