This thesis discusses four novel statistical methods and approximate inference techniques for analyzing structured neural and molecular sequence data. The main contributions are new algorithms for approximate inference and learning in Bayesian latent variable models involving spatial statistics and nonlinear dynamics. First, we propose an amortized variational inference method to separate a set of overlapping signals into spatially localized source functions without knowledge of the original signals or the mixing process. In the second part of this dissertation, we discuss two approaches for uncovering nonlinear, smooth latent dynamics from sequential data. Both algorithms construct variational families on extensions of nonlinear state space models where the underlying systems are described by hidden stochastic differential equations. The first method proposes a structured approximate posterior describing spatially-dependent linear dynamics, as well as an algorithm that relies on the fixed-point iteration method to achieve convergence. The second method proposes a variational backward simulation technique from an unbiased estimate of the marginal likelihood defined through a subsampling process. In the final chapter, we develop connections between discrete and continuous variational sequential search for Bayesian phylogenetic inference. We propose a technique that uses sequential search to construct a variational objective defined on the composite space of non-clock phylogenetic trees. Each of these techniques are motivated by real problems within computational biology and applied to provide insights into the underlying structure of complex data.

Magister Artium - MA / Illegal dumping is a global environmental problem that receives significant management and research attention from various fields of study. Illegally disposed waste (in all of its formats) can cause negative impacts on natural and human environments, and often requires multilevel interventions to abate, or to solve the resultant problems. The impacts of illegal or unsustainable disposal of waste on land, water bodies and the atmosphere contribute to various environmental problems and their levels of intensity at global, regional, national and local scales. Illegal waste disposal also infringes on fundamental human rights that are associated with clean and safe living environments.

Thesis (Ph.D. (Statistics)) -- University of Limpopo, 2021 / In spatial statistics, several methods have been developed to measure the extent of local and global spatial dependence (clustering) in measured data across areas in a region of research interest. These methods are now routinely implemented in most Geographical Information Systems (GIS) and statistical computer packages. However, spatial statistics for measuring joint spatial dependence of multiple spatial measurement and outcome data have not been well developed. A naive analysis would simply apply univariate spatial dependence methods to each data separately. Though this is simple and straightforward, it ignores possible relationships between multiple spatial data because they may be measuring the same phenomena. Limited work has been done on extending the Moran’s index, a commonly used and applied univariate measure of spatial clustering, to bivariate Moran’s index in order to assess spatial dependence for two spatial data. The overall aim of this PhD was to develop multivariate spatial clustering methods for multiple spatial data, especially in the health sciences. Our proposed multivariate spatial clustering statistic is based on the fundamental theory regarding canonical correlations. We firstly reviewed and applied univariate and bivariate Moran’s indexes to spatial analyses of multiple non-communicable diseases and related risk factors in South Africa. Then we derived our proposed multivariate spatial clustering method, which was evaluated by simulation studies and applied to a spatial analysis of multiple non-communicable diseases and related risk factors in South Africa. Simulation studies showed that our proposed multivariate spatial statistic was able to identify correctly clusters of areas with high risks as well as clusters with low risk.

Community health centers (CHCs) provide comprehensive primary care regardless of a patient’s ability to pay. Key policies in the last decade facilitated development of new CHC delivery sites, but the extent of change in geographic access to CHCs (CHC accessibility) is understudied. Furthermore, existing research on CHC accessibility relies on antiquated methods for measuring CHC accessibility despite the growing use of contemporary accessibility methods to study access to other types of health services. We conducted three studies that examined CHC accessibility using contemporary accessibility methods and publicly available data from the Health Resources and Services Administration, American Community Survey, Area Health Resources File, and the 500 Cities Project. The first study assessed CHC accessibility at the census tract level in 2008 and 2016, before and after implementation of policies that expanded CHCs, using the two-step floating catchment area method. It then investigated the association between indicators of CHC need and changes in CHC accessibility. The second study examined the association between CHC accessibility and primary care utilization. The third study investigated differential change in CHC accessibility for census tracts in a subset of Medicaid expansion states compared to census tracts in a subset of non-expansion states. We found that CHC accessibility substantially increased between 2008 and 2016, that spatial distribution of increases in CHC accessibility was not uniform, and that the two-step floating catchment area method could be successfully applied to reveal small area variation in CHC accessibility changes across states. We also found that CHC accessibility was positively related to primary care utilization, but moderated by extent of primary care provider supply and median household income in the surrounding area. Finally, we found that census tracts in Medicaid expansion and non-expansion states saw similar increases in CHC accessibility from 2008 to 2016. In the current health policy context, where gains in health coverage from the Affordable Care Act are in jeopardy of being scaled back, CHC accessibility is critically important. The findings of this work support the important role of CHC accessibility in primary care utilization and describe how CHC accessibility has changed in the last decade. / 2020-10-23T00:00:00Z

Throughout Scandinavia the funeral practices of the Iron Age were, in general, inhumation or cremation. The Iron Age society held many overarching beliefs but with a great ritual flexibility where practices could vary between communities. This study examines the Iron Age burial ground at Nygårds in the parish of Västerhejde, Gotland. The graves consist of both cremated and skeletal remains dating to the period of 400–600 AD. Thus, two different funeral customs were practiced at the burial ground during the same time period. Although excavated in the 1970s, no detailed analyses have been carried out on the Iron Age remains from Nygårds or their grave goods. This study aim is consequently to contribute with new information about the burial ground at Nygårds and the individuals who were interred there. By examining the spatial composition of the two burial customs, their demographic and grave goods, differences and similarities emerge. With the burial grounds spatial structure as a framework, this study discusses the different aspects of the Iron Age life and what the actions of the living can tell us about their dead.

Indiana University-Purdue University Indianapolis (IUPUI) / Research on the association between health and the built environment often delineates environmental exposure using different spatial forms and distances surrounding points of interest, such as residences or schools. Examples from the literature include Euclidian and network buffers, administrative and census boundaries, and other arbitrary geographies, such as grid cells. There is a lack, however, of reports that describe the justifications or implications for using different methods. This research compares different forms and distances for measuring environmental variables surrounding residential locations in the context of adult walking behavior in Marion County, Indiana. Walkability index and vegetation greenness variables were evaluated within 400-meter, 1-kilometer, and 2-kilometer Euclidian and network buffers, census block groups and tracts, and 805- X 805-meter grid cells. Results of analyses using each of these methods to test walkability and greenness as correlates of self-reported walking behavior were compared. Significant differences were observed in measurements of environmental variables as a function of both size and form. There were also significant differences between spatial measure methods when measuring components of walkability and NDVI. Census geographies, widely used in the public health literature, yielded environmental variable measurements differently than did similarly-sized residence-based measure methods. In logistic regressions, the walkability index did not exhibit a significant relationship with self-reported walking behavior. NDVI exhibited a negative relationship with self-reported walking, although the relationship was reversed and significant when stratifying by residential density.

While obesity is now recognized as a major health concern in Ghana, the major drivers, causal factors, and their spatial variation remain unclear. Nutritional changes and lack of physical activity are frequently blamed but the underlying factors, particularly cultural values and practices, remain understudied. Using hot spot analysis and spatial autocorrelation, this research investigates the spatial patterns of obesity in Ghana and the explanatory factors. We also use focus group discussions to examine the primary cultural factors underlying these patterns. The results show that wealth, high education, and urban residence are the best positive predictors of obesity, while poverty, low education, and rural residence are the best (negative) predictors of obesity. Consequently, improving the socioeconomic status, for example, through higher levels of education and urbanization may increase obesity rates. Furthermore, the cultural preference for fat body as the ideal body size drives individual aspiration for weight gain which can lead to obesity. Thus, reducing obesity rates in Ghana is impossible without addressing the underlying cultural values.

Is school accessibility a valued good, and do school closures affect house prices? This thesis applies two different methods of hedonic regression analysis, augmented by spatial regression methods, to a dataset of Hamilton real estate transactions (2005-2017) to investigate whether the closure of a school in an urban neighbourhood negatively affects house prices in that closed school catchment. Evidence is found that school accessibility is a valued good, and that the closure of a primary school will negatively affect house prices from the period of closure announcement through several subsequent quarters. The use of spatial analysis corrects for bias in coefficient estimates. / Thesis / Master of Arts (MA) / Do homebuyers pay a premium to be located closer to a school, and do school closures affect house prices? An analysis of Hamilton real estate transactions (2005-2017) finds evidence that houses closer to schools sell for more, and that a primary school closure has a negative impact on local house sale prices.

No description available.

Food accessibility has lately been of primary interest given its impact on public health outcomes. This thesis illustrates the gaps in food access by applying spatial analysis in Massachusetts accounting for a variety of demographic and socioeconomic factors. The number of grocery stores, farmers markets, and convenience stores within 1/4 and 1 mile of the Census tracts’ centroids are the two accessibility metrics used in the spatial analysis. In addition, a regression model is developed using the Gradient Boosting machine learning method to show the relationship between the socioeconomic factors and the number of grocery stores within 1 mile of the Census tracts’ centroids. Percent of minority population, population in poverty, vehicle ownership, and population density are the factors used as explanatory variables. The results include histograms and maps for the spatial analysis, which show that access to food services is higher in urban high-density areas regardless of high poverty and minority percent values. The regression model results include partial dependence plots, variable importance plot as well as a map that illustrates the standard deviation of the residual values. This research can inform that when the vehicle ownership, the minority and poverty percent increase, the number of grocery stores increases, as these values are high in big cities where there is high availability of food retailers.