Bayesian Inference in Large-scale Problems

Many modern applications fall into the category of "large-scale" statistical problems, in which both the number of observations n and the number of features or parameters p may be large. Many existing methods focus on point estimation, despite the continued relevance of uncertainty quantification in the sciences, where the number of parameters to estimate often exceeds the sample size, despite huge increases in the value of n typically seen in many fields. Thus, the tendency in some areas of industry to dispense with traditional statistical analysis on the basis that "n=all" is of little relevance outside of certain narrow applications. The main result of the Big Data revolution in most fields has instead been to make computation much harder without reducing the importance of uncertainty quantification. Bayesian methods excel at uncertainty quantification, but often scale poorly relative to alternatives. This conflict between the statistical advantages of Bayesian procedures and their substantial computational disadvantages is perhaps the greatest challenge facing modern Bayesian statistics, and is the primary motivation for the work presented here.

Two general strategies for scaling Bayesian inference are considered. The first is the development of methods that lend themselves to faster computation, and the second is design and characterization of computational algorithms that scale better in n or p. In the first instance, the focus is on joint inference outside of the standard problem of multivariate continuous data that has been a major focus of previous theoretical work in this area. In the second area, we pursue strategies for improving the speed of Markov chain Monte Carlo algorithms, and characterizing their performance in large-scale settings. Throughout, the focus is on rigorous theoretical evaluation combined with empirical demonstrations of performance and concordance with the theory.

One topic we consider is modeling the joint distribution of multivariate categorical data, often summarized in a contingency table. Contingency table analysis routinely relies on log-linear models, with latent structure analysis providing a common alternative. Latent structure models lead to a reduced rank tensor factorization of the probability mass function for multivariate categorical data, while log-linear models achieve dimensionality reduction through sparsity. Little is known about the relationship between these notions of dimensionality reduction in the two paradigms. In Chapter 2, we derive several results relating the support of a log-linear model to nonnegative ranks of the associated probability tensor. Motivated by these findings, we propose a new collapsed Tucker class of tensor decompositions, which bridge existing PARAFAC and Tucker decompositions, providing a more flexible framework for parsimoniously characterizing multivariate categorical data. Taking a Bayesian approach to inference, we illustrate empirical advantages of the new decompositions.

Latent class models for the joint distribution of multivariate categorical, such as the PARAFAC decomposition, data play an important role in the analysis of population structure. In this context, the number of latent classes is interpreted as the number of genetically distinct subpopulations of an organism, an important factor in the analysis of evolutionary processes and conservation status. Existing methods focus on point estimates of the number of subpopulations, and lack robust uncertainty quantification. Moreover, whether the number of latent classes in these models is even an identified parameter is an open question. In Chapter 3, we show that when the model is properly specified, the correct number of subpopulations can be recovered almost surely. We then propose an alternative method for estimating the number of latent subpopulations that provides good quantification of uncertainty, and provide a simple procedure for verifying that the proposed method is consistent for the number of subpopulations. The performance of the model in estimating the number of subpopulations and other common population structure inference problems is assessed in simulations and a real data application.

In contingency table analysis, sparse data is frequently encountered for even modest numbers of variables, resulting in non-existence of maximum likelihood estimates. A common solution is to obtain regularized estimates of the parameters of a log-linear model. Bayesian methods provide a coherent approach to regularization, but are often computationally intensive. Conjugate priors ease computational demands, but the conjugate Diaconis--Ylvisaker priors for the parameters of log-linear models do not give rise to closed form credible regions, complicating posterior inference. In Chapter 4 we derive the optimal Gaussian approximation to the posterior for log-linear models with Diaconis--Ylvisaker priors, and provide convergence rate and finite-sample bounds for the Kullback-Leibler divergence between the exact posterior and the optimal Gaussian approximation. We demonstrate empirically in simulations and a real data application that the approximation is highly accurate, even in relatively small samples. The proposed approximation provides a computationally scalable and principled approach to regularized estimation and approximate Bayesian inference for log-linear models.

Another challenging and somewhat non-standard joint modeling problem is inference on tail dependence in stochastic processes. In applications where extreme dependence is of interest, data are almost always time-indexed. Existing methods for inference and modeling in this setting often cluster extreme events or choose window sizes with the goal of preserving temporal information. In Chapter 5, we propose an alternative paradigm for inference on tail dependence in stochastic processes with arbitrary temporal dependence structure in the extremes, based on the idea that the information on strength of tail dependence and the temporal structure in this dependence are both encoded in waiting times between exceedances of high thresholds. We construct a class of time-indexed stochastic processes with tail dependence obtained by endowing the support points in de Haan's spectral representation of max-stable processes with velocities and lifetimes. We extend Smith's model to these max-stable velocity processes and obtain the distribution of waiting times between extreme events at multiple locations. Motivated by this result, a new definition of tail dependence is proposed that is a function of the distribution of waiting times between threshold exceedances, and an inferential framework is constructed for estimating the strength of extremal dependence and quantifying uncertainty in this paradigm. The method is applied to climatological, financial, and electrophysiology data.

The remainder of this thesis focuses on posterior computation by Markov chain Monte Carlo. The Markov Chain Monte Carlo method is the dominant paradigm for posterior computation in Bayesian analysis. It has long been common to control computation time by making approximations to the Markov transition kernel. Comparatively little attention has been paid to convergence and estimation error in these approximating Markov Chains. In Chapter 6, we propose a framework for assessing when to use approximations in MCMC algorithms, and how much error in the transition kernel should be tolerated to obtain optimal estimation performance with respect to a specified loss function and computational budget. The results require only ergodicity of the exact kernel and control of the kernel approximation accuracy. The theoretical framework is applied to approximations based on random subsets of data, low-rank approximations of Gaussian processes, and a novel approximating Markov chain for discrete mixture models.

Data augmentation Gibbs samplers are arguably the most popular class of algorithm for approximately sampling from the posterior distribution for the parameters of generalized linear models. The truncated Normal and Polya-Gamma data augmentation samplers are standard examples for probit and logit links, respectively. Motivated by an important problem in quantitative advertising, in Chapter 7 we consider the application of these algorithms to modeling rare events. We show that when the sample size is large but the observed number of successes is small, these data augmentation samplers mix very slowly, with a spectral gap that converges to zero at a rate at least proportional to the reciprocal of the square root of the sample size up to a log factor. In simulation studies, moderate sample sizes result in high autocorrelations and small effective sample sizes. Similar empirical results are observed for related data augmentation samplers for multinomial logit and probit models. When applied to a real quantitative advertising dataset, the data augmentation samplers mix very poorly. Conversely, Hamiltonian Monte Carlo and a type of independence chain Metropolis algorithm show good mixing on the same dataset.

Pitit se rìches malere [Children are the wealth of the poor]: The Influence of Gender and Power on Choice and Uptake of Long Acting Contraceptives

Background: Haiti has the highest maternal mortality rate in the Latin American and Caribbean region. Despite the fact that Haiti has received twice as much family planning assistance as any other country in the western hemisphere, the unmet need for contraception remains particularly high. Our hypothesis is that unsuccessful efforts of family planning programs may be related to a misconstrued understanding of the complex role of gender in relationships and community in Haiti. This manuscript is one of four parts of a study that intends to examine some of these issues with a particular focus on the influence of uptake and adherence to long acting contraceptive (LAC) methods.

Methods: We conducted a three-month community-based qualitative assessment through 20 in-depth interviews in Fondwa, Haiti. Participants were divided into 4 groups of five: female users, female non-users, men and key community stakeholders.

Results: Based on the qualitative interviews, we found that main barriers included lack of access to family planning education and services and concerns regarding side effects and health risks, especially related to menstrual disruption and fears of infertility. Women have a constant pressure to remain fertile and bear children, due not only to social but also economic needs. As relationships are conceived as means for economic provision, the likelihood of uptake of irreversible methods (vasectomy and tubal ligation) was restricted by loss of fertility. Consequently, the discourse of family planning, though self-recognized in their favor, assumes women can afford not to bear children. This assumption should be questioned given the complexities of the other social determinants at play, all which affect the reproductive decisions made by Haitians.

Conclusions: Overall, our study indicated awareness surrounding contraception in the Haitian Fondwa community. Combining the substantial impact of birth spacing with the elevated yet unmet need for contraceptives in the area, it is necessary to address the intricacies of gender issues in order to implement successful programing. In Haiti not being able to bear a child poses a threat to economic and social survival, possibly explaining a dimension of the low uptake of LACs in the region, even when made available. For this reason, we believe IUDs (Intrauterine Devices) provide a suitable alternative, allowing the couple to comprehend all of the factors involved in decision making, thus decreasing the imbalances of power and knowledge prior to considering an irreversible alternative.

Enhancing the Visualization of the Peripheral Retina with Wide Field-of-View Optical Coherence Tomography

The goal of my Ph.D. thesis is to enhance the visualization of the peripheral retina using wide-field optical coherence tomography (OCT) in a clinical setting.

OCT has gain widespread adoption in clinical ophthalmology due to its ability to visualize the diseases of the macula and central retina in three-dimensions, however, clinical OCT has a limited field-of-view of 300. There has been increasing interest to obtain high-resolution images outside of this narrow field-of-view, because three-dimensional imaging of the peripheral retina may prove to be important in the early detection of neurodegenerative diseases, such as Alzheimer's and dementia, and the monitoring of known ocular diseases, such as diabetic retinopathy, retinal vein occlusions, and choroid masses.

Before attempting to build a wide-field OCT system, we need to better understand the peripheral optics of the human eye. Shack-Hartmann wavefront sensors are commonly used tools for measuring the optical imperfections of the eye, but their acquisition speed is limited by their underlying camera hardware. The first aim of my thesis research is to create a fast method of ocular wavefront sensing such that we can measure the wavefront aberrations at numerous points across a wide visual field. In order to address aim one, we will develop a sparse Zernike reconstruction technique (SPARZER) that will enable Shack-Hartmann wavefront sensors to use as little as 1/10th of the data that would normally be required for an accurate wavefront reading. If less data needs to be acquired, then we can increase the speed at which wavefronts can be recorded.

For my second aim, we will create a sophisticated optical model that reproduces the measured aberrations of the human eye. If we know how the average eye's optics distort light, then we can engineer ophthalmic imaging systems that preemptively cancel inherent ocular aberrations. This invention will help the retinal imaging community to design systems that are capable of acquiring high resolution images across a wide visual field. The proposed model eye is also of interest to the field of vision science as it aids in the study of how anatomy affects visual performance in the peripheral retina.

Using the optical model from aim two, we will design and reduce to practice a clinical OCT system that is capable of imaging a large (800) field-of-view with enhanced visualization of the peripheral retina. A key aspect of this third and final aim is to make the imaging system compatible with standard clinical practices. To this end, we will incorporate sensorless adaptive optics in order to correct the inter- and intra- patient variability in ophthalmic aberrations. Sensorless adaptive optics will improve both the brightness (signal) and clarity (resolution) of features in the peripheral retina without affecting the size of the imaging system.

The proposed work should not only be a noteworthy contribution to the ophthalmic and engineering communities, but it should strengthen our existing collaborations with the Duke Eye Center by advancing their capability to diagnose pathologies of the peripheral retinal.