Abstract:Summary Generalised linear mixed regression models are fundamental in statistics. Modelling random effects that are shared by individuals allows for correlation among those individuals. There are many methods and statistical packages available for analysing data using these models. Most require some form of numerical or analytic approximation because the likelihood function generally involves intractable integrals over the latents. The Bayesian approach avoids this issue by iteratively sampling the full conditional distributions for various blocks of parameters and latent random effects. Depending on the choice of the prior, some full conditionals are recognisable while others are not. In this paper we develop a novel normal approximation for the random effects full conditional, establish its asymptotic correctness and evaluate how well it performs. We make the case for hierarchical binomial and Poisson regression models with canonical link functions, for hierarchical gamma regression models with log link and for other cases. We also develop what we term a sufficient reduction (SR) approach to the Markov Chain Monte Carlo algorithm that allows for making inferences about all model parameters by replacing the full conditional for the latent variables with a considerably reduced dimensional function of the latents. We expect that this approximation could be quite useful in situations where there are a very large number of latent effects, which may be occurring in an increasingly 'Big Data' world. In the sequel, we compare our methods with INLA, which is a particularly popular method and which has been shown to be excellent in terms of speed and accuracy across a variety of settings. Our methods appear to be comparable to theirs in terms of accuracy, while INLA was faster, for the settings we considered. In addition, we note that our methods and those of others that involve Gibbs sampling trivially handle parameters that are functions of multiple parameters, while INLA approximations do not. Our primary illustration is for a three‐level hierarchical binomial regression model for data on health outcomes for patients who are clustered within physicians who are clustered within particular hospitals or hospital systems.

A comment on the "A unified Bayesian inference framework for generalized linear models"

A Unified Bayesian Inference Framework for Generalized Linear Models

An AMP-Based Low Complexity Generalized Sparse Bayesian Learning Algorithm

Comment on An and Schorfheide's Bayesian Analysis of DSGE Models

Generalised Bayes Linear Inference

Scalable Bayesian inference for the generalized linear mixed model

Advances in Bayesian model selection consistency for high-dimensional generalized linear models

Bayesian views of generalized additive modelling

Using a linear mixed-effect model framework to estimate multivariate generalizability theory parameters in R

Bayesian Nonlinear Regression using Sums of Simple Functions

General Design Bayesian Generalized Linear Mixed Models

Generalized Memory Approximate Message Passing for Generalized Linear Model.

On the maximum likelihood estimation in general log-linear models

Estimation for High-Dimensional Multi-Layer Generalized Linear Model -- Part II: The ML-GAMP Estimator

Inference in generalized bilinear models

Approximate inferences for Bayesian hierarchical generalised linear regression models

Quasi-Monte Carlo Estimation in Generalized Linear Mixed Models

A Multiscale Perspective on Maximum Marginal Likelihood Estimation

R-VGAL: a sequential variational Bayes algorithm for generalised linear mixed models

The Bridged Posterior: Optimization, Profile Likelihood and a New Approach to Generalized Bayes

A Variational Approach for Modeling High-dimensional Spatial Generalized Linear Mixed Models