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.

Revisiting a destination: A tourist’s choice model : A case study of Sweden

This paper documents the design and results of a study on tourists’ decision-making about destinations in Sweden. For this purpose, secondary data, available from surveys were used to identify which type of individual has the highest probability to revisit a destination and what are influencing factors to do so. A binary logit model is applied. The results show that very important influencing factors are the length of stay as well as the origin of the individual. These results could be useful for a marketing organization as well as for policy, to develop strategies to attract the most profitable tourism segment. Therefore, it can also be a great support for sustainable tourism development, where the main focus does not has priority on increasing number of tourists.

Detecting Free Riders in Collective Brands through a Hierarchical Choice Process

Free-riding behaviors exist in tourism and they should be analyzed from a comprehensive perspective; while the literature has mainly focused on free riders operating in a destination, the destinations themselves might also free ride when they are under the umbrella of a collective brand. The objective of this article is to detect potential free-riding destinations by estimating the contribution of the different individual destinations to their collective brands, from the point of view of consumer perception. We argue that these individual contributions can be better understood by reflecting the various stages that tourists follow to reach their final decision. A hierarchical choice process is proposed in which the following choices are nested (not independent): “whether to buy,” “what collective brand to buy,” and “what individual brand to buy.” A Mixed Logit model confirms this sequence, which permits estimation of individual contributions and detection of free riders.

Determining the season of death from the family composition of insects infesting carrion

Determining the season of death by means of the composition of the families of insects infesting carrion is rarely attempted in forensic studies and has never been statistically modelled. For this reason, a baseline-category logit model is proposed for predicting the season of death as a function of whether the area where the carcass was exposed is sunlit or shaded and of the relative abundance of particular families of carrion insects (Calliphoridae, Fanniidae, Sarcophagidae, and Formicidae). The field study was conducted using rodent carcasses (20-252 g) in an urban forest in southeastern Brazil. Four carcasses (2 in a sunlit and 2 in a shaded area) were placed simultaneously at the study site, twice during each season from August 2003 through June 2004. The feasibility of the model, measured in terms of overall accuracy, is 64 +/- 14%. It is likely the proposed model will assist forensic teams in predicting the season of death in tropical ecosystems, without the need of identifying the species of specimens or the remains of carrion insects.

Predictors of non-response to a questionnaire survey of a volunteer twin panel: Findings from the Australian 1989 twin cohort

Questionnaire surveys, while more economical, typically achieve poorer response rates than interview surveys. We used data from a national volunteer cohort of young adult twins, who were scheduled for assessment by questionnaire in 1989 and by interview in 1996-2000, to identify predictors of questionnaire non-response. Out of a total of 8536 twins, 5058 completed the questionnaire survey (59% response rate), and 6255 completed a telephone interview survey conducted a decade later (73% response rate). Multinomial logit models were fitted to the interview data to identify socioeconomic, psychiatric and health behavior correlates of non-response in the earlier questionnaire survey. Male gender, education below University level, and being a dizygotic rather than monozygotic twin, all predicted reduced likelihood of participating in the questionnaire survey. Associations between questionnaire response status and psychiatric history and health behavior variables were modest, with history of alcohol dependence and childhood conduct disorder predicting decreased probability of returning a questionnaire, and history of smoking and heavy drinking more weakly associated with non-response. Body-mass index showed no association with questionnaire non-response. Despite a poor response rate to the self-report questionnaire survey, we found only limited sampling biases for most variables. While not appropriate for studies where socioeconomic variables are critical, it appears that survey by questionnaire, with questionnaire administration by telephone to non-responders, will represent a viable strategy for gene-mapping studies requiring that large numbers of relatives be screened.

The host-tourist interaction in a world heritage site: the case of Guimarães

Guimarães, in the northwest of Portugal, is a city of strong symbolic and cultural significance and its nomination by UNESCO as world heritage, in 2001, enlarged its tourism potential. In this paper we present a few results of a survey that envisaged capturing the Guimarães residents’ perceptions of tourism impacts and their attitudes towards tourists. Specifically, one analyzes the type of relationship that exists between some socio-demographic groups and the perceived tourism impacts, as well as their socio-characteristics and the existing level of interaction between residents and tourists. The survey was implemented between January and March 2010 to a convenience sample of 540 inhabitants of the municipality of Guimarães resulting in 400 questionnaires with complete data. For this, we made use of various statistical techniques. Using a factorial analysis, we can conclude that the three factors used explain 52.3% of the variance contained in the original variables obtained from the survey. By another side, using a logit model in the analysis and taking as the dependent variable the frequent or very frequent contact with tourists, we found that only the variables referred to perceived positive impacts of tourism, education and the place of residence in urban areas have shown to be statistically significant. We are aware of the multiple ways the issue of residents’ perceptions and attitudes towards tourism can be approached and of the difficulties to get useful policy-oriented insights. This paper is a step in that trail.

A produção de educação nos Açores: a nota de matemática

Tese de Doutoramento, Ciências Económicas e Empresariais (Desenvolvimento Económico e Social e Economia Pública), 16 de Janeiro de 2014, Universidade dos Açores.

Estudo empírico sobre insolvências no setor de alojamento e restauração

Mestrado em Contabilidade e Análise Financeira

Determinantes do estado de saúde dos Portugueses

Dissertação apresentada como requisito parcial para obtenção do grau de Mestre em Estatística e Gestão de Informação

Fossil fuel dependence and interfuel substitution in the electricity sector of Western European countries between 1990-2011

A Work Project, presented as part of the requirements for the Award of a Masters Degree in Economics from the NOVA – School of Business and Economics

Consumer satisfaction with water, wastewater and waste services in Portugal

While the concept of consumer satisfaction is a central topic in modern marketing theory and practice, citizens' satisfaction with public services, and especially water and waste services, is a eld that still remains empirically rather unexplored. The following study aims to contribute to this area by analysing the determinants of user satisfaction in the water, wastewater and waste sector in Portugal, using a unique survey of 1070 consumers undertaken by the Portuguese Water and Waste Regulator ERSAR. I perform an analysis of the relation between overall service satisfaction and attributespeci c service satisfaction with an ordered logit model. I then explore if subjective consumer satisfaction can be re ected by ERSAR's technical performance indicators. The results suggest that overall consumer satisfaction is driven by consumer's satisfaction with speci c service aspects but unrelated to socioeconomic and demographic characteristics. Furthermore, I show that there is no monotonic association between ERSAR's technical performance indicators and consumers' levels of satisfaction.

A Salience Theory of Choice Errors

We study a psychologically based foundation for choice errors. The decision maker applies a preference ranking after forming a 'consideration set' prior to choosing an alternative. Membership of the consideration set is determined both by the alternative specific salience and by the rationality of the agent (his general propensity to consider all alternatives). The model turns out to include a logit formulation as a special case. In general, it has a rich set of implications both for exogenous parameters and for a situation in which alternatives can a¤ect their own salience (salience games). Such implications are relevant to assess the link between 'revealed' preferences and 'true' preferences: for example, less rational agents may paradoxically express their preference through choice more truthfully than more rational agents.

Stochastic Choice and Consideration Sets

We model a boundedly rational agent who suffers from limited attention. The agent considers each feasible alternative with a given (unobservable) probability, the attention parameter, and then chooses the alternative that maximises a preference relation within the set of considered alternatives. We show that this random choice rule is the only one for which the impact of removing an alternative on the choice probability of any other alternative is asymmetric and menu independent. Both the preference relation and the attention parameters are identi fied uniquely by stochastic choice data.

Patients' and physicians' preferences for type 2 diabetes mellitus treatments in Spain and Portugal: a discrete choice experiment.

OBJECTIVE To assess Spanish and Portuguese patients' and physicians' preferences regarding type 2 diabetes mellitus (T2DM) treatments and the monthly willingness to pay (WTP) to gain benefits or avoid side effects. METHODS An observational, multicenter, exploratory study focused on routine clinical practice in Spain and Portugal. Physicians were recruited from multiple hospitals and outpatient clinics, while patients were recruited from eleven centers operating in the public health care system in different autonomous communities in Spain and Portugal. Preferences were measured via a discrete choice experiment by rating multiple T2DM medication attributes. Data were analyzed using the conditional logit model. RESULTS Three-hundred and thirty (n=330) patients (49.7% female; mean age 62.4 [SD: 10.3] years, mean T2DM duration 13.9 [8.2] years, mean body mass index 32.5 [6.8] kg/m(2), 41.8% received oral + injected medication, 40.3% received oral, and 17.6% injected treatments) and 221 physicians from Spain and Portugal (62% female; mean age 41.9 [SD: 10.5] years, 33.5% endocrinologists, 66.5% primary-care doctors) participated. Patients valued avoiding a gain in bodyweight of 3 kg/6 months (WTP: €68.14 [95% confidence interval: 54.55-85.08]) the most, followed by avoiding one hypoglycemic event/month (WTP: €54.80 [23.29-82.26]). Physicians valued avoiding one hypoglycemia/week (WTP: €287.18 [95% confidence interval: 160.31-1,387.21]) the most, followed by avoiding a 3 kg/6 months gain in bodyweight and decreasing cardiovascular risk (WTP: €166.87 [88.63-843.09] and €154.30 [98.13-434.19], respectively). Physicians and patients were willing to pay €125.92 (73.30-622.75) and €24.28 (18.41-30.31), respectively, to avoid a 1% increase in glycated hemoglobin, and €143.30 (73.39-543.62) and €42.74 (23.89-61.77) to avoid nausea. CONCLUSION Both patients and physicians in Spain and Portugal are willing to pay for the health benefits associated with improved diabetes treatment, the most important being to avoid hypoglycemia and gaining weight. Decreased cardiovascular risk and weight reduction became the third most valued attributes for physicians and patients, respectively.

The Impacts of social networks on immigrants' employment prospect: the spanish case 1997-2007

This paper studies the extent to which social networks influence the employment stability and wages of immigrants in Spain. By doing so, I consider an aspect that has not been previously addressed in the empirical literature, namely the connection between immigrants' social networks and labor market outcomes in Spain. For this purpose, I use micro-data from the National Immigrant Survey carried out in 2007. The analysis is conducted in two stages. First, the impact of social networks on the probability of keeping the first job obtained in Spain is studied through a multinomial logit regression. Second, quantile regressions are used to estimate a wage equation. The empirical results suggest that once the endogeneity problem has been accounted for, immigrants' social networks influence their labor market outcomes. On arrival, immigrants experience a mismatch in the labor market. In addition, different effects of social networks on wages by gender and wage distribution are found. While contacts on arrival and informal job access mechanisms positively influence women's wages, a wage penalty is observed for men.