We consider predictive inference using a class of temporally dependent Dirichlet processes driven by Fleming–Viot diffusions, which have a natural bearing in Bayesian nonparametrics and lend the resulting family of random probability measures to analytical posterior analysis. Formulating the implied statistical model as a hidden Markov model, we fully describe the predictive distribution induced by these Fleming–Viot-driven dependent Dirichlet processes, for a sequence of observations collected at a certain time given another set of draws collected at several previous times. This is identified as a mixture of Pólya urns, whereby the observations can be values from the baseline distribution or copies of previous draws collected at the same time as in the usual Pólya urn, or can be sampled from a random subset of the data collected at previous times. We characterize the time-dependent weights of the mixture which select such subsets and discuss the asymptotic regimes. We describe the induced partition by means of a Chinese restaurant process metaphor with a conveyor belt, whereby new customers who do not sit at an occupied table open a new table by picking a dish either from the baseline distribution or from a time-varying offer available on the conveyor belt. We lay out explicit algorithms for exact and approximate posterior sampling of both observations and partitions, and illustrate our results on predictive problems with synthetic and real data.

Approximate Bayesian Computation (ABC) methods are increasingly used for inference in situations in which the likelihood function is either computationally costly or intractable to evaluate. Extensions of the basic ABC rejection algorithm have improved the computational efficiency of the procedure and broadened its applicability. The ABC – Population Monte Carlo (ABC-PMC) approach has become a popular choice for approximate sampling from the posterior. ABC-PMC is a sequential sampler with an iteratively decreasing value of the tolerance, which specifies how close the simulated data need to be to the real data for acceptance. We propose a method for adaptively selecting a sequence of tolerances that improves the computational efficiency of the algorithm over other common techniques. In addition we define a stopping rule as a by-product of the adaptation procedure, which assists in automating termination of sampling. The proposed automatic ABC-PMC algorithm can be easily implemented and we present several examples demonstrating its benefits in terms of computational efficiency.

Two-dimensional (2D) nuclear magnetic resonance (nmr) methods have become increasingly popular in metabolomics, since they have considerable potential to accurately identify and quantify metabolites within complex biological samples. 2D ${}^1$H J-resolved (jres) nmr spectroscopy is a widely used method that expands overlapping resonances into a second dimension. However, existing analytical processing methods do not fully exploit the information in the jres spectrum and, more importantly, do not provide measures of uncertainty associated with the estimates of quantities of interest, such as metabolite concentration. Combining the data-generating mechanisms and the extensive prior knowledge available in online databases, we develop a Bayesian method to analyse 2D jres data, which allows for automatic deconvolution, identification and quantification of metabolites. The model extends and improves previous work on one-dimensional nmr spectral data. Our approach is based on a combination of B-spline tight wavelet frames and theoretical templates, and thus enables the automatic incorporation of expert knowledge within the inferential framework. Posterior inference is performed through specially devised Markov chain Monte Carlo methods. We demonstrate the performance of our approach via analyses of datasets from serum and urine, showing the advantages of our proposed approach in terms of identification and quantification of metabolites.

For time-ordered functional data, an important yet challenging task is to forecast functional observations with uncertainty quantification. Scalar predictors are often observed concurrently with functional data and provide valuable information about the dynamics of the functional time series. We develop a fully Bayesian framework for dynamic functional regression, which employs scalar predictors to model the time-evolution of functional data. Functional within-curve dependence is modeled using unknown basis functions, which are learned from the data. The unknown basis provides substantial dimension reduction, which is essential for scalable computing, and may incorporate prior knowledge such as smoothness or periodicity. The dynamics of the time-ordered functional data are specified using a time-varying parameter regression model in which the effects of the scalar predictors evolve over time. To guard against overfitting, we design shrinkage priors that regularize irrelevant predictors and shrink toward time-invariance. Simulation studies decisively confirm the utility of these modeling and prior choices. Posterior inference is available via a customized Gibbs sampler, which offers unrivaled scalability for Bayesian dynamic functional regression. The methodology is applied to model and forecast yield curves using macroeconomic predictors, and demonstrates exceptional forecasting accuracy and uncertainty quantification over the span of four decades.

A hierarchical Bayesian factor model for multivariate spatially and temporally correlated data is proposed. This method searches factor scores incorporating a dependence within observations due to both a geographical and a temporal structure and it is an extension of a model proposed by Mezzetti (2012) using the results of a separable covariance matrix for the spatial panel data as in Leorato and Mezzetti (2016). A Gibbs sampling algorithm is implemented to sample from the posterior distributions. We illustrate the benefit and the performance of our model by analyzing death rates for different diseases together with some socio-economical and behavioural indicators and by analyzing simulated data.

A multiple changepoint model in continuous time is formulated as a continuous-time hidden Markov model, defined on a countable infinite state space. The new formulation of the multiple changepoint model allows the model complexities, i.e. the number of changepoints, to accrue unboundedly upon the arrivals of new data. Inference on the number of changepoints and their locations is based on a collapsed Gibbs sampler. We suggest a new version of forward-filtering backward-sampling (FFBS) algorithm in continuous time for simulating the full trajectory of the latent Markov chain, i.e. the changepoints. The FFBS algorithm is based on a randomized time-discretization for the latent Markov chain through uniformization schemes, combined with a discrete-time version of FFBS algorithm. It is shown that, desirably, both the computational cost and the memory cost of the FFBS algorithm are only quadratic to the number of changepoints. The new formulation of the multiple changepoint models allows varying scale of run lengths of changepoints to be characterized. We demonstrate the methods through simulations and a real data example for earthquakes.

We establish Bayesian effect selection for the broad class of structured additive distributional regression models using a spike and slab prior specification with scaled beta prime marginals for the importance parameters of blocks of regression coefficients. This enables us to model and select effects in all distributional parameters, such as location, scale, skewness or correlation parameters, for arbitrary distributions. The regression specifications encompass various effect types such as non-linear or spatial effects. Our spike and slab prior relies on a parameter expansion that separates blocks of regression coefficients into overall scalar importance parameters and vectors of standardised coefficients, and yields effective shrinkage and good sampling performance. Using constrained priors, it is possible to implement effect decompositions, where, for example, a non-linear effect can be decomposed into a linear component and the non-linear deviation from this linear effect; and to select both separately. We investigate some shrinkage properties, propose a way of eliciting prior hyperparameters and provide full posterior inference through Markov Chain Monte Carlo simulations. Using both simulated and real data sets, we show that our approach is applicable for data with various functional covariate effects, multilevel predictors and non-standard response distributions, such as bivariate Gaussian or zero-inflated Poisson.

Parameter inference for stochastic differential equation mixed effects models (SDEMEMs) is challenging. Analytical solutions for these models are rarely available, which means that the likelihood is also intractable. In this case, exact inference (up to the discretisation of the stochastic differential equation) is possible using particle MCMC methods. Although the exact posterior is targeted by these methods, a naive implementation for SDEMEMs can be highly inefficient. Our article develops three extensions to the naive approach which exploit specific aspects of SDEMEMs and other advances such as correlated pseudo-marginal methods. We compare these methods on simulated data and data from a tumour xenography study on mice.

With modern high-dimensional data, complex statistical models are necessary, requiring computationally feasible inference schemes. We introduce Max-and-Smooth, an approximate Bayesian inference scheme for a flexible class of latent Gaussian models (LGMs) where one or more of the likelihood parameters are modeled by latent additive Gaussian processes. Our proposed inference scheme is a two-step approach. In the first step (Max), the likelihood function is approximated by a Gaussian density with mean and covariance equal to either (a) the maximum likelihood estimate and the inverse observed information, respectively, or (b) the mean and covariance of the normalized likelihood function. In the second step (Smooth), the latent parameters and hyperparameters are inferred and smoothed with the approximated likelihood function. The proposed method ensures that the uncertainty from the first step is correctly propagated to the second step. Because the prior density for the latent parameters is assumed to be Gaussian and the approximated likelihood function is Gaussian, the approximate posterior density of the latent parameters (conditional on the hyperparameters) is also Gaussian, thus facilitating efficient posterior inference in high dimensions. Furthermore, the approximate marginal posterior distribution of the hyperparameters is tractable, and as a result, the hyperparameters can be sampled independently of the latent parameters. We show that the computational cost of Max-and-Smooth is close to being insensitive to the number of independent data replicates, and that it scales well with increased dimension of the latent parameter vector provided that its Gaussian prior density is specified with a sparse precision matrix. In the case of a large number of independent data replicates, sparse precision matrices, and high-dimensional latent vectors, the speedup is substantial in comparison to an MCMC scheme that infers the posterior density from the exact likelihood function. The accuracy of the Gaussian approximation to the likelihood function increases with the number of data replicates per latent model parameter. The proposed inference scheme is demonstrated on one spatially referenced real dataset and on simulated data mimicking spatial, temporal, and spatio-temporal inference problems. Our results show that Max-and-Smooth is accurate and fast.

We introduce an approach based on the Givens representation for posterior inference in statistical models with orthogonal matrix parameters, such as factor models and probabilistic principal component analysis (PPCA). We show how the Givens representation can be used to develop practical methods for transforming densities over the Stiefel manifold into densities over subsets of Euclidean space. We show how to deal with issues arising from the topology of the Stiefel manifold and how to inexpensively compute the change-of-measure terms. We introduce an auxiliary parameter approach that limits the impact of topological issues. We provide both analysis of our methods and numerical examples demonstrating the effectiveness of the approach. We also discuss how our Givens representation can be used to define general classes of distributions over the space of orthogonal matrices. We then give demonstrations on several examples showing how the Givens approach performs in practice in comparison with other methods.

Markov chain Monte Carlo is a key computational tool in Bayesian statistics, but it can be challenging to monitor the convergence of an iterative stochastic algorithm. In this paper we show that the convergence diagnostic $\hat{R}$ of Gelman and Rubin (1992) has serious flaws. Traditional $\hat{R}$ will fail to correctly diagnose convergence failures when the chain has a heavy tail or when the variance varies across the chains. In this paper we propose an alternative rank-based diagnostic that fixes these problems. We also introduce a collection of quantile-based local efficiency measures, along with a practical approach for computing Monte Carlo error estimates for quantiles. We suggest that common trace plots should be replaced with rank plots from multiple chains. Finally, we give recommendations for how these methods should be used in practice.