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.

The impact of widowhood on Irish mortality due to suicide and accidents

The impact of widowhood on suicide and accident mortality in Ireland was investigated using Poisson regression analysis applied to routine data relating to all 10 561 suicidal and accidental deaths of married or widowed persons aged at least 35 years in Ireland during 1986–2005. Mortality rates were almost always higher among the widowed and often by a 2-fold, statistically significant difference. The excess mortality was equivalent to 2083 or 57.6% of all suicidal or accidental deaths of widowed persons in 1986–2005. Routine contact with recently widowed persons by public health professionals may be warranted with a view to reducing their excess mortality.

Life Experiences of Youth Who Were Born with HIV Infection in Puerto Rico: The Voices of Young Survivors

The effectiveness of antiretroviral therapy (ART) transformed the pediatric HIV epidemic. The disease changed significantly over the course of three decades: while early in the epidemic it was almost always fatal, it has become a chronic condition. This study examined how perinatally-infected youth experience the impact of HIV in their lives. A qualitative study using interpretative phenomenological analysis (IPA) was conducted. Twenty in-depth interviews were carried out among 12 women and 8 men aged 18 to 30 years in Puerto Rico. These were conducted in Spanish, audio-recorded, transcribed and translated into English. While narrating their experiences, participants were interpreting what the situation meant to them and how they make sense of it. Three topics emerged: (1) perception and response to treatment and illness, particularly their lived experiences with ART; (2) disclosure experiences; and (3) family matters. Most participants challenged their therapy, in most cases to force their caregivers to disclose their status. Problems with adherence were attributed to busy schedules or forgetfulness. Participants experienced the disfiguring adverse effects of ART, which they endured for years without being informed that ART was the cause of these. Participants’ experiences with disclosure demonstrated the importance of validating them as individuals capable of managing their health. The paternalistic approach of withholding their diagnosis to spare them suffering resulted in increased anxiety. Participants acknowledged the difficulties of revealing their HIV status to their partners. They referred to family and friends as essential in coping with HIV. However, some encountered discrimination and stigma within their families. Participants who had suffered the loss of their parents found other parental figures such as adoptive parents or other family members. Most participants expressed a desire to have children. Perinatally HIV-infected youth will require health services for the rest of their lives. The adult health care into which they transition should consider their needs and journey. Services should consider including family members. This study underscores the need for improved access to mental health services. It is also essential to transcend medical treatment and develop a broader perspective of health care. Health care services should include reproductive decision-making counselling services.

Particle flux in the NE Atlantic as recorded by sediment traps

Downward particle flux was measured using sediment traps at various depths over the Porcupine Abyssal Plain (water depth ab. 4850 m) for prolonged periods from 1989 to 1999. A strong seasonal pattern of flux was evident reaching a maximum in mid-summer. The composition of the material changed with depth, reflecting the processes of remineralisation and dissolution as the material sank through the water column. However, there was surprisingly little seasonal variation in its composition to reflect changes in the biology of the euphotic zone. Currents at the site have a strong tidal component with speeds almost always less than 15 cm/sec. In the deeper part of the water column they tend to be northerly in direction, when averaged over periods of several months. A model of upper ocean biogeochemistry forced by meteorology was run for the decade in order to provide an estimate of flux at 3000 m depth. Agreement with measured organic carbon flux is good, both in terms of the timings of the annual peaks and in the integrated annual flux. Interannual variations in the integrated flux are of similar magnitude for both the model output and sediment trap measurements, but there is no significant relationship between these two sets of estimates. No long-term trend in flux is evident, either from the model, or from the measurements. During two spring/summer periods, the marine snow concentration in the water column was assessed by time-lapse photography and showed a strong peak at the start of the downward pulse of material at 3000 m. This emphasises the importance of large particles during periods of maximum flux and at the start of flux peaks. Time lapse photographs of the seabed show a seasonal cycle of coverage of phytodetrital material, in agreement with the model output both in terms of timing and magnitude of coverage prior to 1996. However, after a change in the structure of the benthic community in 1996 no phytodetritus was evident on the seabed. The model output shows only a single peak in flux each year, whereas the measured data usually indicated a double peak. It is concluded that the observed double peak may be a reflection of lowered sediment trap efficiency when flux is very high and is dominated by large marine snow particles. Resuspension into the trap 100 m above the seabed, when compared to the primary flux at 3000 m depth (1800 mab) was lower during periods of high primary flux probably because of a reduction in the height of resuspension when the material is fresh. At 2 mab, the picture is more complex with resuspension being enhanced during the periods of higher flux in 1997, which is consistent with this hypothesis. However there was rather little relationship to flux at 3000 m in 1998. At 3000 m depth, the Flux Stability Index (FSI), which provides a measure of the constancy of the seasonal cycle of flux, exhibited an inverse relationship with flux, such that the highest flux of organic carbon was recorded during the year with the greatest seasonal variation.

Labour and Low Energy Buildings: the energy performance gap as Social Practice

Buildings are responsible for approximately 30% of EU end-use emissions (Bettgenhäuser , et al, 2009) and are at the forefront of efforts to meet emissions targets arising from their design, construction and operation. For the first time in its history, construction industry outputs must meet specific energy targets if planned reductions in greenhouse gas emissions are to be achieved through nearly zero energy buildings (nZEB) (EC, 2010) supported by on-site renewable heat and power. Where individual UK dwellings have been tested before occupation to assess whether they meet energy design criteria, the results indicate what is described as an ‘energy performance gap’, that is, energy use is almost always more than that specified. This leads to the conclusion that the performance gap is, inter alia, a function of the labour process and thus a function of social practice. Social practice theory, based on Schatzki’s model (2002), is utilised to explore the performance gap as a result of the changes demanded in the social practice of building initiated by new energy efficiency rules. The paper aims to open a discussion where failure in technical performance is addressed as a social phenomenon.

Measuring Scope of Practice Enactment among Primary Care Registered Nurses in Ontario

Background: Primary care is the sector of health care in which patients first establish contact with the health system, are provided person-focused care over time for all new or common needs, and receive coordinated integrated health services provided elsewhere by other members of the health care team. Registered Nurses (RNs) in Canada provide care within this sector in varying roles. The extent to which RNs enact their full scope of practice in primary care settings in Canada is not known. The Actual Scope of Practice questionnaire (ASCOP) is a 26 item Likert scale questionnaire developed by researchers in Canada and validated in the acute care setting to measure the extent to which RNs apply the knowledge, skills and competencies of the professional full scope of practice. Similar to the acute care setting, there is a need to measure scope of practice enactment in the primary care setting. Objectives: The overall aim of this thesis was to measure scope of practice enactment in the primary care setting. Two research objectives were addressed: (1) to revise and adapt the ASCOP questionnaire for use in the primary care setting, and (2) to determine internal consistency, construct validity, and sensitivity of the modified instrument, the ASCOP-PC. Methods: To address the first objective, a narrative literature review and synthesis and an expert panel review was conducted. To address the second objective a cross-sectional survey of 178 RNs working in primary care organizations in Ontario was conducted Results: The ASCOP, with few modifications, addressed key attributes of nursing scope of practice in the primary care setting. The ASCOP-PC yielded acceptable alpha coefficients ranging from 0.66 to 0.91 and explained variances from 44.2 to 62.6. Total mean score of 5.16 suggests that RNs within these models of care almost always engage in activities reflected in the ASCOP-PC. Interpretation: Findings from this study support the use of a the modified ASCOP questionnaire as a reliable and valid measure of scope of practice enactment among primary care nurses in the primary care setting.

"Per aurem intrat Christus in Mariam". Aproximación iconográfica a la "conceptio per aurem" en la pintura italiana del Trecento desde fuentes patrísticas y teológicas

La tesis mariológica de la conceptio per aurem, según la cual la Virgen María habría concebido a Jesucristo por el oído en el momento de escuchar del ángel el mensaje celestial anunciándole que, sin perder su virginidad, sería madre del Hijo de Dios encarnado, ha merecido hasta ahora muy pocos estudios académicos rigurosamente fundados en fuentes primarias. De hecho, en la literatura especializada son muy escasas las referencias a tal teoría y, cuando algún estudioso la evoca, casi siempre se contenta con aludir a ella, sin aportar pruebas documentales. Sin embargo, tal como lo revelan las nueve pinturas italianas aquí analizadas, esa teoría fue ilustrada mediante sutiles metáforas visuales en muchas obras pictóricas medievales, las cuales se inspiraron en una sólida tradición literaria. Además una pléyade de Padres de la Iglesia y teólogos medievales testimonia, mediante afirmaciones explícitas, que semejante teoría gozó de notable aceptación entre los maestros del pensamiento cristiano. Basándose en numerosos textos patrísticos y teológicos, este artículo intenta dos objetivos esenciales: exponer, ante todo, las distintas formulaciones teóricas propuestas por esos pensadores; y además, tratar de poner en luz los significados dogmáticos que subyacen bajo esa sorprendente tesis.

Desistance from Sexual Offending: Do the Mainstream Theories Apply?

The literature on desistance from crime has become well established in recent years with strong bodies of evidence supporting the role of factors such as employment, relationships and identity change in this process. However, the relevance of this literature to individuals convicted of sexual crimes is not known as such individuals are almost always excluded from this research. This article presents the results from one of the first empirical studies on desistance from sexual offending based on 32 in-depth life story interviews with adult males previously convicted of child sex offences. In this analysis we explore the significance of work, the role of relationships, and changes in imagined selves in the self-identities of individuals successfully desisting from sexual offending. The findings provide support for all three factors in helping to sustain desistance from sex offending, but also suggest clear differences between desistance from sex offending and other types of crime in these regards.

Licenshantering inom webbprojekt : Hur medvetna är företag om tredjepartslicenser?

Teknikutvecklingen går snabbt framåt, och idag finns det ett stort behov av att använda andra utvecklares kod för att hänga med i det höga tempot. De kallas samlat för ramverk eller bibliotek, och hjälper utvecklaren att på ett effektivare sätt ta sig från start till mål utan att behöva skriva all programmeringskod själv. Dessa tredjepartslösningar är nästintill alltid bundna till ett licensavtal, vars restriktioner och tillåtelser utvecklaren måste följa vid nyttjandet. I denna studie har vi undersökt hur medvetenheten ser ut kring de licenser som är bundna till dessa tredjepartslösningar. Då det framkom i vår förstudie att vårt fall hade en relativt låg medvetenhet har vi även valt att titta på hur medvetenheten kan ökas. För att genomföra detta har vi valt att intervjua utvecklare och projektledare på ett konsultföretag. Vi undersökte även vilka faktorer som är viktiga för att höja medvetenheten samt vilka konsekvenser som kunde uppkomma vid bristfällande licenshantering. Vi upptäckte att det var en bristfällig kunskap om tredjepartslicenser på det studerade företaget, och hur de följde de licensrestriktioner som fanns för respektive licens. För att höja medvetenheten föreslår vi hjälpmedel i form av en automatiserad centraliserad lösning, lathundar för en enklare överblick av licensavtalen och hur användning av redan färdiga programvaror kan hjälpa till att öka medvetenheten och hanteringen av licenser.

Practical calculations for uniform external pressure buckling in cylindrical shells with stepped walls

Metal cylindrical storage structures of significant size, such as silos and vertical-axis tanks, are almost always constructed from many short cylindrical shells of different thickness as the stress resultants on the wall progressively increase towards the base. The resulting increases in thickness are always made in step changes using metal sheets of uniform thickness because of the availability of such source materials. The result is a shell with a stepped wall with multiple discrete steps in thickness. Such shells are very susceptible to buckling under external pressure when empty or partially filled, but the buckling mode may involve only part of the shell height due to the changes in shell thickness. These changes must therefore be accounted for within the design process. A new method of determining the critical buckling resistance of such shells was recently developed, and although it has been shown to be valid, the methodology for its application in practical design has not been set out or shown. This paper therefore briefly describes the new method and demonstrates the manner in which it can be used to produce rapid, safe assessments of cylindrical shells with a wide range of patterns of wall thickness changes. The results are then suitable for direct introduction into such documents as the European standard on metal shells [1] and the ECCS Recommendations [2].

Meio ambiente e planejamento: a relação cidade-natureza nos Planos Urbanísticos da cidade de Natal no século XX

This research studies the relation between city and nature in the urbanity s plans from Natal RN, Brazil, during the 20th century. Investigates and analyses the main documents that bring environmental s ideas inside from the urbanity s plans of Natal; gives the historical and economical situation from the city in each period studied; features the urbanity s plans, using categories of analyses to comprehension of this instruments. Try to contribute in the bigger process of historical rescue of Natal, and stimulates new studies. It was used documental s research, and bibliographic material. It was identified four (04) kinds of plans: the ones that focus in health and aesthetic (1901, 1929, 1935) technology and science (1968), zoning and control (1974, 1984) and environment. The hints founded shows that environmental ideas were put inside of the plans by government demands, especially in 1994 s plan, almost always without popular contribution and without this population get understands its meanings and implications

An exploratory study of orthoptic student satisfaction in a blended learning environment

Purpose: To describe orthoptic student satisfaction in a blended learning environment. Methods: Blended learning and teaching approaches that include a mix of sessions with elearning are being used since 2011/2012 involving final year (4th year) students from an orthoptic program. This approach is used in the module of research in orthoptics during the 1 semester. Students experienced different teaching approaches, which include seminars, tutorial group discussions and e-learning activities using the moodle platform. The Constructivist OnLine Learning Environment Survey (COLLES ) was applied at the end of the semester with 24 questions grouped in 6 dimensions with 4 items each: Relevance to professional practice, Reflection, Interactivity, Tutor support, Peer support and Interpretation. A 5-point Likert scale was used to score each individual item of the questionnaire (1 - almost never to 5 – almost always). The sum of items in each dimension ranged between 4 (negative perception) and 20 (positive perception). Results: Twenty-four students replied to the questionnaire. Positive points were related with Relevance (16.13±2.63), Reflection (16.46±2.45), Tutor support (16.29±2.10) and Interpretation (15.38±2.16). The majority of the students (n=18; 75%) think that the on-line learning is relevant to students’ professional practice. Critical reflections about learning contents were frequent (n=19; 79.17%). The tutor was able to stimulate critical thinking (n=21; 87.50%), encouraged students to participate (n=18; 75%) and understood well the student’s contributions (n=15; 62.50%). Less positive points were related with Interactivity (14.13±2.77) and Peer support (13.29±2.60). Response from the colleagues to ideas (n=11; 45.83%) and valorization of individual contributions (n=10; 41.67%) scored lower than other items. Conclusions: The flow back and forth between face-to-face and online learning situations helps the students to make critical reflections. The majority of the students are satisfied with a blended e-learning system environment. However, more work needs to be done to improve interactivity and peer support.

Multiple endocrine neoplasia, the old and the new: a mini review

Multiple endocrine neoplasia syndromes have since been classified as types 1 and 2, each with specific phenotypic patterns. MEN1 is usually associated with pituitary, parathyroid and paraneoplastic neuroendocrine tumours. The hallmark of MEN2 is a very high lifetime risk of developing medullary thyroid carcinoma (MTC) more than 95% in untreated patients. Three clinical subtypesdMEN2A, MEN2B, and familial MTC (FMTC) have been defined based on the risk of pheochromocytoma, hyperparathyroidism, and the presence or absence of characteristic physical features). MEN2 occurs as a result of germline activating missense mutations of the RET (REarranged during Transfection) proto-oncogene. MEN2-associated mutations are almost always located in exons 10, 11, or 13 through 16. Strong genotype-phenotype correlations exist with respect to clinical subtype, age at onset, and aggressiveness of MTC in MEN2. These are used to determine the age at which prophylactic thyroidectomy should occur and whether screening for pheochromocytoma or hyperparathyroidism is necessary. Specific RET mutations can also impact management in patients presenting with apparently sporadic MTC. Therefore, genetic testing should be performed before surgical intervention in all patients diagnosed with MTC. Recently, Pellegata et al. have reported that germline mutations in CDKN1B can predispose to the development of multiple endocrine tumours in both rats and humans and this new MEN syndrome is named MENX and MEN4, respectively. CDKN1B. A recent report showed that in sporadic MTC, CDKN1B V109G polymorphism correlates with a more favorable disease progression than the wild-type allele and might be considered a new promising prognostic marker. New insights on MEN syndrome pathogenesis and related inherited endocrine disorders are of particular interest for an adequate surgical and therapeutic approach.

Primary neuroendocrine lung tumor presenting with acute ileal obstruction. Case report

The authors describe a clinical case of a patient with neuroendocrine carcinoma of the lung diagnosed after the onset of an intestinal obstruction from an ileal metastasis. A review of literature reveals that the incidence of symptomatic gastro-intestinal metastases from lung cancer has been estimated to be about 2-3% and is exceedingly rare that the intestinal symptoms may be the initial presentation of cancer of the lung. The authors emphasize the difficulty of preoperative diagnosis of gastro-intestinal metastases which is made, almost always, too late because of the lack of specific symptoms. In our case, on account of the computed tomography, we leaned towards the diagnosis of lymphoma because of the double mediastinal and abdominal localization. Furthermore, this diagnosis was supported by the fact that the pulmonary lesion did not have clear radiological features of a lung cancer. The prognosis is poor because once intestinal metastases occur, other metastatic sites, which would make surgery only a palliative measure, are already present. The review of the literature shows that the average survival rate of these patients is 136 days. In our case the patient survived 277 days.

Whiteness and news: the interlocking social construction of "realities"

This dissertation examines how mainstream U.S. journalism consistently serves white racial interests and the racial status quo, or what I call white incumbency, despite its push for diversity and its stated aims to improve coverage of nonwhite communities. It is based on an in-depth ethnographic study of two daily newspapers and extensive one-on-one interviews with 61 journalists. I found that although journalists strongly identify with the need for more diverse coverage in newspapers, they emphasize individual and personal stories that avoid recognition of historical racial power imbalances, exhibiting what Ruth Frankenberg calls power-evasive race cognizance. Journalists also demonstrate a number of often contradictory identifications and self-understandings about themselves and their work, such as commitments to diversity and not taking sides, but these conflicts are almost always resolved in favor of white incumbency. Journalistic conventions and practices, such as the watchdog function and its emphasis on public institutions, routinely produce stories that replay and reinforce racial hegemony by portraying nonwhites as problems or people seeking “special privileges.” Also, journalistic repertoires about those conventions and practices avoid interrogations of journalists’ ongoing complicity in the maintenance of white incumbency.