Fitting latent non-Gaussian models using variational Bayes and Laplace approximations

TL;DR

This paper develops fast variational Bayes algorithms for latent non-Gaussian models, enhancing robustness and applicability in various spatial and temporal data analysis contexts, and introduces an accessible R package extension.

Contribution

It introduces scalable variational Bayes methods for LnGMs, enabling robust inference and providing an easy-to-use R package extension from LGMs.

Findings

The algorithms effectively downweight extreme events in latent processes.

The methods are applicable to diverse models like autoregressive and spatial models.

The ngvb package simplifies extending LGMs to LnGMs in R.

Abstract

Latent Gaussian models (LGMs) are perhaps the most commonly used class of models in statistical applications. Nevertheless, in areas ranging from longitudinal studies in biostatistics to geostatistics, it is easy to find datasets that contain inherently non-Gaussian features, such as sudden jumps or spikes, that adversely affect the inferences and predictions made from an LGM. These datasets require more general latent non-Gaussian models (LnGMs) that can handle these non-Gaussian features automatically. However, fast implementation and easy-to-use software are lacking, which prevent LnGMs from becoming widely applicable. In this paper, we derive variational Bayes algorithms for fast and scalable inference of LnGMs. The approximation leads to an LGM that downweights extreme events in the latent process, reducing their impact and leading to more robust inferences. It can be applied to a wide range of models, such as autoregressive processes for time series, simultaneous autoregressive models for areal data, and spatial Mat\'ern models. To facilitate Bayesian inference, we introduce the ngvb package, where LGMs implemented in R-INLA can be easily extended to LnGMs by adding a single line of code.