Approximate inference for disease mapping with sparse Gaussian processes

Abstract

Gaussian process (GP) models are widely used in disease mapping as they provide a natural framework for modeling spatial correlations. Their challenges, however, lie in computational burden and memory requirements. In disease mapping models, the other difficulty is inference, which is analytically intractable due to the non-Gaussian observation model. In this paper, we address both these challenges. We show how to efficiently build fully and partially independent conditional (FIC/PIC) sparse approximations for the GP in two-dimensional surface, and how to conduct approximate inference using expectation propagation (EP) algorithm and Laplace approximation (LA). We also propose to combine FIC with a compactly supported covariance function to construct a computationally efficient additive model that can model long and short length-scale spatial correlations simultaneously. The benefit of these approximations is computational. The sparse GPs speed up the computations and reduce the memory requirements. The posterior inference via EP and Laplace approximation is much faster and is practically as accurate as via Markov chain Monte Carlo.

Statistical Analysis of Spatial Expression Pattern for Spatially Resolved Transcriptomic Studies

“…GLSM, also known as the generalized linear geostatistical model or the spatial generalized linear mixed model, is a generalized linear mixed model that directly models non-Gaussian spatial data and uses random effects to capture the underlying stationary spatial process. GSLM has been commonly used for interpolating and prediction of spatial data, with applications in spatial disease mapping and spatial epidemiologic studies 66,67 . However, different from all these previous GLSM development, we instead focus on developing a hypothesis testing framework for GLSM.…”

Sparse Spatio-temporal Gaussian Processes with General Likelihoods

“…Since the covariance of this prior is of same form as in sparse GPs, we can use same tricks as presented, e.g., in [20] to speed up the inference. With this we achieve the overall complexity O(N T nm 2 ), where N is the number of EP iterations across the time sequence (in our examples we used N = 3, which we empirically observed to be sufficient).…”

Fast and flexible linear mixed models for genome-wide genetics

“…i=1 g i (n 3 + p i n 2 )) Present Work the highest ML (or REML) score for each marker to compute approximate likelihood 53 ratio and Wald tests [32,38], or analogously derive posterior distributions and Bayes 54 factors by summing appropriate statistics across the grid. The Grid-LMM approach 55 relies on a re-parameterization of the typical LMM framework from individual variance 56 components σ 2 l to variance component proportions h 2 l = σ 2 l /σ 2 , where σ 2 without the 57 subscript denotes the total sum of all variance components (including the residual). 58 Since the variance components must be non-negative, their proportions are restricted to 59 the unit interval [0, 1] and sum to 1, forming a simplex.…”