.png){kind=icon}
.png){kind=icon}
Return
Cutting Feedback and Modularized Analyses in Generalized Bayesian Inference
.png){kind=icon}
PRE
AI
DOI：10.1214/24-BA1448.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This work considers Bayesian inference under misspecification for complex statistical models comprised of simpler submodels, referred to as modules, that are coupled together. Such multi-modular models often arise when combining information from different data sources where there is a module for each data source. When some of the modules are misspecified, the challenges of Bayesian inference under misspecification can sometimes be addressed by using cutting feedback methods, which modify conventional Bayesian inference by limiting the influence of unreliable modules. Here we investigate cutting feedback methods in the context of generalized posterior distributions built from loss functions. We make three main contributions. First, we describe how cutting feedback methods can be defined in the generalized Bayes setting, and discuss the appropriate scaling of the loss functions in this context. Second, we derive a novel type of conditional Laplace approximation that accurately describes the behavior of the posterior for a given module's parameters when we condition on parameters in other modules. Third, we leverage this novel result to provide several convenient diagnostics for Bayesian modular inference, which we then apply to examples in the literature on cut model inference.
Keywords:
cutting feedback
model misspecification
modularization
semi-modular inference
generalized Bayesian inference
Journal
IF:
2.5
Papers:
34
Citations:
3.0K