Bayesian model averaging
Definition
Bayesian model averaging is a statistical approach used in Bayesian analysis of graphical models that accounts for uncertainty in network structure and parameters when testing conditional independence between variables. Within network psychometrics, the inclusion Bayes factor—a method based on Bayesian model averaging techniques—allows researchers to test for the presence or absence of edges in Markov Random Field models by specifying prior distributions for both the network structure and edge weight parameters. The method addresses limitations by quantifying uncertainty about network parameters and structure. The sensitivity of the inclusion Bayes factor to prior specifications—particularly the scale of priors on edge weights—is critical for drawing robust inferences about conditional independence relationships.
Sources: Sekulovski et al. (2024)
Related Terms
Applications
Bayesian Model Averaging and Conditional Independence Testing
Bayesian model averaging provides the theoretical foundation for the inclusion Bayes factor, which is used to test conditional independence between pairs of variables in network psychometric models. This approach allows researchers to account for uncertainty about network structure when determining whether two variables are directly connected after controlling for the remaining variables in the network.
Sources: Sekulovski et al. (2024)
Bayesian Model Averaging and Prior Distributions
The application of Bayesian model averaging in network psychometrics requires specification of prior distributions for the network structure and edge weight parameters. The sensitivity of the inclusion Bayes factor to these prior choices—particularly variations in the scale of the prior on partial correlations—substantially affects the method's ability to distinguish between the presence and absence of edges.
Sources: Sekulovski et al. (2024)
