Evidence linking microbiome to human health is rapidly growing, suggesting that the human microbiome may serve as novel biomarkers for disease. In microbiome profiling, the direct sequencing outputs are counts or compositions of bacterial taxa at different taxonomic levels. The key research questions in microbiome analysis are to identify the bacterial taxa associated with a clinical outcome and to leverage microbiome aiming to make an accurate prediction on host phenotypes. Although the next generation sequencing has produced extensive microbiome data, data analyses are hindered by several statistical challenges due to the unique characteristics of microbiome profile which includes that 1) the number of taxa greatly exceeds the sample size, 2) most taxa are in extremely low abundance and absent in many samples, and 3) the taxa are related to one another by an evolutionary tree. In this dissertation, three papers are presented to address these challenges. In the first paper, a regularized variance component model is developed for selecting important microbiome taxa. We consider regression analysis by treating bacterial taxa at different levels as multiple random effects. For each taxon, a kernel matrix is calculated based on distance measures in the phylogenetic tree and it acts as one variance component in the joint model. Then, taxonomic selection is achieved by the lasso (least absolute shrinkage and selection operator) penalty on variance components. Our method integrates biological information into the variable selection problem and greatly improves selection accuracies. Simulation studies demonstrate the superiority of our methods versus existing ones, for example, the group-lasso. This method is then applied to a longitudinal microbiome study of Human Immunodeficiency Virus (HIV) infected patients. We implement our method using the high performance computing language Julia. Software and detailed documentation are freely available at https://github.com/JingZhai63/VCselection. In the second paper, we thoroughly investigate the link between lung microbiome composition, pulmonary inflammation, and early lung dysfunction. First, the genus-level taxa are labeled as two pneumotypes: supraglottic and background predominant taxa (i.e., SPT and BPT as previously described by others). Next, multiple statistical approaches are used to characterize these two pneumotype taxa, including dissimilarity-overlap curve (DOC) microbiome dynamics analysis, weighted Spearman correlation analysis, network analysis, etc. We find that previously defined microbial taxa have different effect on inflammation markers and lung function in an HIV positive population. The dynamics of post-ART pneumotype SPT was host-dependent, indicating that the microbiome variability among individuals not only originates from the difference in microbiome species assemblage, but also stems from host specific factors, such as lifestyles. The complex lung function-microbiome-inflammatory network and microbiome dynamics pattern suggest that a healthy lung microbiome may play a critical role in preventing lung function decline of HIV infected individuals. The findings in the second paper suggest that, in addition to identify taxa related to disease, it is also important to uncover the microbiome-phenotype network by understanding microbiome as a whole. In the third paper, we present DeepBiome, a deep learning model, to uncover the network of microbiome and visualize its path to disease. The proposed DeepBiome takes microbiome abundance data as input and uses the phylogenetic tree as a prior knowledge to decide the optimal number of layers and neurons on each of it. By doing so, we are able to relieve the computation burden of tuning DNN hyperparameters. In addition, DeepBiome provides phylogeny regularized weight decay to improve prediction on host phenotypes during training. This deep learning framework not only can analyze a microbiome as a whole to provide a comprehensive network view and but also can identify taxa associate with outcome at each taxonomic level. DeepBiome is designed for both regression and classification problems to support a broader application in microbiome analysis. The simulation studies and real data application show that DeepBiome is a cutting-edge tool in the area of complex microbiome data analysis.

This dissertation revisits the variable subject pronoun expression (SPE) in Spanish from an understudied perspective: written narratives. Based on data collected among monolingual speakers (L1), second language learners (L2), and heritage language learners (HL) of Spanish, the present study aims to further our understanding of SPE’s behaviour in written language. As the breadth of the existent research indicates, the variability of subject pronoun expression (SPE) in spoken Spanish (overt vs. null) is of great interest. Recently, the span of this research has moved beyond the expression of subjects among native speakers (e.g. Lastra & Martín Butragueño, 2015; Michnowicz, 2015) to explore its behaviour among L2 learners (Geeslin et al., 2013, 2015; Linford, 2016), and HL (Bessett, 2017; Cacoullos & Travis, 2015). While SPE varies in terms of overall rates of expression across Spanish dialects, a similar underlying grammar has been found to predict its behavior. Variationist research on SPE among L1 speakers of Spanish has identified factors such as tense, mood, and aspect (TMA) of the verb, TMA continuity, switch reference, as well as person and number of the referent as key predictors of its realization (Carvalho et al., 2015, p. x-xi). The current study explores whether these underlying patterns are also found in written narratives, a register that is different from spoken language (Biber & Conrad, 2009). A total of 216 essays narrating past events were collected among L1, L2, and HL participants (45 compositions in each group), resulting in a total of 5,075 tokens of 1st person SPE. Due to the heterogeneity in proficiency levels that is usually found in L2 and HL classrooms, the compositions were analyzed based on accuracy levels represented by three parameters: morphological, syntactical, and lexical accuracy. Each participant was then regrouped and classified as beginner, intermediate, or advanced. The analysis revealed that written and spoken SPE behave similarly in terms of the linguistic predictors that impact pronominal usage among L1 speakers. In addition, it was possible to identify a gradual acquisition of this variable among L2 and HL learners across accuracy levels. Finally, the results clearly show that written production triggers fewer overt subject pronouns and experiences a desensitization to switch-reference contexts when compared to spoken language. By bridging the fields of second language acquisition and sociolinguistic variation, this research deepens our understanding about L1 variation in written language and sheds light on the differences among L2 and HL learners across accuracy levels.

In this dissertation, I investigate the role of household ritual in the development of complex society at the Maya site of Ceibal, in Guatemala. My approach to ritual – and its role in both uniting communities and fostering hierarchies – is grounded in practice theory, particularly the work of Catherine Bell. Ceibal was the ideal location at which to explore the relationship between domestic and public ritual during the development of Maya civilization, because the site was founded as a ceremonial center around the transition to sedentism and intensive agriculture. My dissertation research consisted of four seasons of extensive excavations at the Karinel Group, an early residential area at Ceibal, along with laboratory analyses in Guatemala City and bibliographic research into other early Maya sites. The Karinel Group was selected based on the presence of early ceramics, proximity to the site center, and relatively shallow overburden. It was important to develop a precise chronology of the Karinel Group, in order to learn 1) when people began to live in permanent patio groups, occupied over generations; 2) when household rituals and public plaza rituals were similar, and when they differed; and 3) when ancestor veneration rituals, as documented in northern Belize, began to play a role in the development of social inequality. The chronology was based on careful analyses of stratigraphy, ceramics, and radiocarbon dates. I found that while the Karinel Group was occupied around 850-775 BC, the earliest permanent structures date to after 775 BC and the earliest patio group dates to after 700 BC. These results support other research at Ceibal that indicates the transition from a more mobile lifestyle to a more sedentary one was gradual and piecemeal. I also found that domestic and public rituals at Ceibal had little in common until about 350 BC. Contrary to a common assumption, public rituals do not necessarily develop out of domestic predecessors. At Ceibal, separate household rituals centered on low circular platforms and involving ceramic figurines may have existed in tension with increasingly hierarchical rituals involving greenstone axe caches and restricted platforms in the public plaza. Interestingly, early ancestor veneration rituals seen in northern Belize were absent at Ceibal and in reports from other parts of the Maya lowlands, suggesting that the social processes that shaped early Maya society differed across regions. Around 350 BC, a major shift occurred across the lowlands, when household rituals became more centralized and similar to public rituals. This transformation in ritual practices requires further investigation, but it may have paved the way for the emergence of rulers.

Down syndrome (DS) is the most common genetic form of intellectual disability. Individuals with DS present memory and learning difficulties associated with hippocampal impairments. The studies presented in this dissertation investigated different ways of encoding that may rely less on the hippocampus and therefore support better long-term retention for individuals with DS compared to typically developing (TD) mental age-match children. Study 1 investigated two learning conditions where participants learned novel arbitrary picture word associations using fast mapping (FM), an incidental, exclusion-based learning procedure and an explicit encoding learning procedure (Coutanche & Thompson-Schill, 2014). Both DS and TD participants performed similarly for the two conditions across immediate, 5-minute, and 1-week delay. The DS group was not impaired compared to the TD group and maintained what they had learned over the 1-week delay. These findings could have been due to receiving multiple test opportunities at each delay test. Therefore, Study 2 investigated three learning conditions where participants either learned a series of action sequences from two demonstrations and a pause, three demonstrations, or from two demonstrations and a test. The DS group benefited from the test condition at the 1-month delay test. Further comparisons between sequences tested vs. not tested at the 5-minute delay test revealed a testing benefit in both groups. The benefit seen with testing could be influence by a self-bias benefit when given the opportunity to model the action sequences themselves vs. watching another person model the action sequences. Therefore, Study 3 investigated the self-reference effect (SRE), which is a memory advantage for information encoded in reference to one’s self compared to another person or a lower level processing condition. This study consisted of three learning conditions where participants sorted toys and made toy evaluations in reference to themselves, an experimenter, or based on the size of the toy. There was an overall benefit of the size evaluation condition and no difference between the DS and TD group for the self and other condition at immediate or 1-week recognition and source test. Task difficulty could have inhibited SRE in this study. These findings suggest that retrieval practice could be an alternative method of learning that could be beneficial for individuals with DS. Future studies can investigate the benefit of expanded retrieval practice (i.e., combination of spacing and testing). Additionally, more work is needed to better determine the outcome of other ways of learning (i.e., FM and SR) for young children and individuals with DS. Memory interventions and learning strategies can incorporate retrieval practice to support memory difficulties that individuals with DS display.

Background: Prediabetes and diabetes are major public health problems. Diabetes is one of the ten leading causes of death in Arizona. The Centers for Disease Control and Prevention (CDC) report that at least 84 million U.S. adults (33.9%) ages 18 and older have prediabetes or at risk for type 2 diabetes. The CDC also reports that Medicaid beneficiaries are among the group at greatest risk for prediabetes. The National Diabetes Prevention Program (DPP), a widely accepted evidence-based program aims at reducing the risk of type 2 diabetes through lifestyle modifications and behavior change. Although the DPP has been proven to delay the onset of type 2 diabetes, very few states in the United States have adopted diabetes prevention policies to offer the DPP as a covered benefit for Medicaid beneficiaries. Given the multiplicity of factors that underpin the support and enactment of public health policies, researchers must take an integrated approach to understanding the data-driving decisions of legislators and policymakers. The overall objective of this dissertation is to generate the necessary data to inform diabetes prevention policy in Arizona. We hypothesize that Arizona legislators and key policymakers are likely to support diabetes prevention policies that will reduce the burden of disease, are cost-saving, and align with constituency personal narratives and values. To test this hypothesis, the following four aims will be explored in this dissertation: Aim 1: Estimate the prevalence of prediabetes in the state of Arizona using commercial laboratory diagnostic data from 2012-2017; Aim 2: Project the economic net-cost savings of a diabetes prevention intervention in Arizona; Aim 3: Conduct a case study on the best practices from the states of Vermont, Minnesota, California, and Montana with successful diabetes prevention coverage policies; and Aim 4: Understand high-level processes of decision-making concerning diabetes prevention through the lens of the Arizona State Legislature. Methods: Employing the National Academy of Sciences, Assessing Prevalence and Trends (ATP) framework, prediabetes age-adjusted (2000 U.S. Standard Population) prevalence rates were calculated using 2012-2017 commercial laboratory data from Sonora Quest Laboratories, LLC. (SQL). Prediabetes cases were selected if they met one of the clinical criteria for prediabetes as defined by the American Diabetes Association 2018 Standards of Medical Care in Diabetes: HbA1C = 5.7-6.4%, fasting glucose = 100-125 mg/dl; or oral glucose tolerance test (OGTT) = 140-199 mg/dl. To eliminate prediabetes cases with multiple laboratory screenings, cases were selected based on patient’s most recent laboratory test. Chi-square (χ2) tests were performed to test for statistically significant differences in age-adjusted prevalence of prediabetes by sex. Data were managed and analyzed in SAS 9.4. Applying a theoretical cost model, the economic cost-savings for the DPP were estimated using the prediabetes age-adjusted prevalence estimates, net-costs of the DPP and the total medical expenditures for those with and without a DPP intervention in Arizona as input indicators. James E. Anderson five-stage conceptual policymaking framework (1. Policy identification, 2. Policy formulation, 3. Policy adoption, 4. Policy implementation, and 5. Policy evaluation) guided the case study methodology of this dissertation. Key informants (N=10) from the Diabetes Prevention and Control and Medicaid programs from the states of Vermont (n=2), Minnesota (n=4), California (n=2), and Montana (n=2) were recruited via email and telephone for qualitative in-depth interviews. Interviews were transcribed and coded for thematic policy analysis and analyzed in NVivo-10. A 10-item mixed-survey was constructed in Qualtrics® and administered to members of the Arizona State Legislators (N=90; 30 Senators and 60 House of Representatives) via email and in-person. Survey responses were analyzed in Qualtrics®. Results: For those referred for prediabetes testing to SQL, the aggregate 2012-2017 age-adjusted prediabetes prevalence in Arizona was 19.7% (95% CI 19.6-19.8), 21.8% (95% CI 21.7-21.9) for men, and 18.2% (95% CI 18.2-18.3) for women. Analysis using χ2 test determined a statistically significant difference in age-adjusted prediabetes prevalence by sex (p<0.001). Out of 380,702 individuals that would benefit from a DPP intervention, a theoretical cost model estimated approximately $4.3 billion in medical costs-savings, and $3.6 billion net cost-savings at year 1 of the DPP compared to those with no DPP intervention. Over a ten-year period, the net cost-savings were estimated at $4.2 billion. Based on Anderson’s framework for policy process, four themes emerged through the key-informants: 1. Strategic partnerships, 2. Federal and state financial support for DPP implementation, 3. Institutional leadership support by Health and Medicaid Departments, and 4. Health care and reimbursement payment models of the DPP. The Arizona State Legislature sought advocacy groups, certified diabetes educators, and legislature leadership as integral to supporting evidence-based diabetes prevention policies. Advocacy groups were identified as an ‘extremely influential’ factor, while fiscal impact was deemed as a ‘very important’ factor whether a diabetes related bill passes or fails within a given legislative session. Conclusions: The mining of commercial diagnostic laboratory data may be powerful to enhancing existing surveillance systems that estimate the prevalence of prediabetes. The DPP in Arizona is shown to produce medical-savings and overall net-costs savings to the economy of Arizona. The enactment of diabetes prevention policies at the state level are complexed and multifaceted. Policymakers may generate informed decisions concerning diabetes prevention policies in Arizona through an array of stakeholders comprised of advocacy groups and coalitions, while also supporting evidence-based policies that are fiscally conservative and that address the social determinants of public health in relation to diabetes. Policy recommendations: Evidence-based public health policies are key in preventing the rise of type 2 diabetes in Arizona and warrants attention from policymakers as well as diabetes advocacy groups and coalitions. It is imperative that Arizona’s Medicaid system supports the implementation of a reimbursement/coverage policy for the DPP to ensure the sustainability of evidence-based prevention programs to help curb the incidence of type 2 diabetes, in addition to utilizing large commercial laboratory diagnostic data to estimate the burden of prediabetes in Arizona.

This study utilizes historic chronic disease and employment data from the University of Michigan Health and Retirement Study (HRS) to specify a multivariate regression model for forecasting the impact of specific chronic diseases on work life. The purpose of the study is to create a new model for forecasting chronic disease-adjusted work life years, which is a measure of disease burden due to chronic disease. The ability to work may be considered a proxy for quality of life, as it is a means by which a person preserves their financial independence and maintains their financial capacity for self-care. This forecasting model is germane to advanced nursing practice, as it provides practitioners a tool to measure patients’ ability to work given various scenarios of chronic disease – many of which, are preventable. This tool may be useful for motivating patients to adopt healthy lifxestyle behaviors such as smoking cessation, weight loss, exercise, and adopting healthy eating habits so they may change chronic disease trajectories and preserve their ability to work and financially provide for themselves and their families. This advocacy and promotion of patient health through self-care is a cornerstone of advanced nursing practice (Thrasher, 2002). Furthermore, this tool may also be useful for calculating or forecasting disease burden in terms of an individual’s attenuated work years or lost productivity. On a larger scale, this tool may be used to calculate lost labor force participation of a population or group of individuals. These statistics may be used as quality improvement measures, economic forecasting data, or for justifying healthcare policy changes or for the allocation of healthcare resources.

Teacher assessment literacy is essential to the success of teaching (T. H. Wang et al., 2008), the quality of student learning (DeLuca, Klinger, Searle, & Shulha, 2010; Mertler, 2004; White, 2009) and student learning motivation (Alkharusi, 2013; Dorman & Knightley, 2006). However, studies have consistently shown inadequate levels of assessment literacy for both pre-service and in-service teachers in language teaching and general education (Alkharusi, 2011; Malone, 2013; Mertler & Campbel1, 2005; Plake & Impara, 1992). This dissertation research aims to understand the current state of EFL teacher assessment literacy in six colleges of applied sciences in Oman by investigating teacher self-perceived language assessment literacy (competence and frequency of practice) as well as assessment knowledge and practice in the classroom. Its theoretical framework are Brookhart’s (2011) contemporary conceptualizations of formative assessment, Eyal’s (2012) discussion of digital assessment literacy as an important component in measuring teacher assessment literacy and Alkharusi’s (2009; 2010) methodological approach to investigating assessment literacy. This study implements a mixed-method approach using a triangulation of five data sources including self-assessment surveys, a language assessment knowledge test, an assessment evaluation task, classroom observations with a focus on teacher-produced assessment tasks, and teacher interviews. Multivariate analyses were used to examine relations between assessment literacy and participant background characteristics, such as gender, academic preparation, teaching experience, experience as coordinators, pre-service training, and in-service training. Research findings point to strengths and weaknesses in EFL teacher assessment literacy as well as matches and mismatches between teachers’ self-perceived assessment literacy and demonstrated assessment knowledge. Overall EFL teachers view themselves as moderately competent in language assessment. The variable with the strongest effect on teacher’s assessment literacy was pre-service training in assessment. The study discusses implications of research findings and offers suggestions for EFL teacher preparation programs, teaching institutions, and future research.

This dissertation, “We Are This People and We Intend to Endure As Such”: Black and Indigenous Peoplehood and Persistence, engages the Indigenous peoplehood matrix, a core theoretical construct developed by Indigenous scholars and comprised of four interrelated elements: land, language, sacred history, and ceremonial cycle. In a comprehensive overview of the peoplehood matrix, I note that the model emerges, in part, from anthropology’s theories of enduring peoples and persistent cultural systems, then modified and further fortified with Indigenous epistemologies. In my research, I employ the peoplehood matrix to delineate Indigenous cultural identity and introduce a model of African American peoplehood, respectfully adapted from the Indigenous peoplehood matrix, comprised of the elements of embodied Blackness, expression, heritage, and spirituality. I underscore the attributes of both matrices that represent persistent cultural identity systems. Through this analysis, I demonstrate that, in both models, the interconnectedness of the elements of peoplehood establishes and maintains persistent cultural identity with a built-in capacity, in the elements of sacred history and heritage, for continuity—a necessary characteristic for persistence, particularly within the milieu of settler colonialism. Within this context, I argue that resistance to assimilation and erasure further coalesces a people, is inherent to American Indian and African American peoplehood, and corresponds with continuing demands for sovereignty and emancipation. In making this argument, I identify appropriation and annihilation as the double-pronged settler-colonial imperative in relation to American Indigeneity and American Blackness, while also naming various mechanisms of the peoples’ ongoing resistance to it. Here, I introduce and elucidate the ideas of “oppositional identification,” “oppositional coalescence,” and “contrast mechanisms” and advance the Peoplehood and Persistence model, a tool for the analysis of boundary maintenance. Further, I examine the settler-biopolitical aims to exploit Black bodies and American Indian lands via strategic educational policies and the eradication of American Indian gender systems, many with extra-binary gender roles.

The presented three studies attempted to enhance our understanding of the evolution of stony hillslope in drylands. Multi-temporal terrestrial LiDAR scans were conducted at six positions around stony plots (2 m × 6.1 m) with three slope treatments (5%, 12%, and 20%, two replications for each slope treatment) through 22 simulated rainfall applications. The first study examines the potential use of terrestrial LiDAR scanner to measure erosion on small plots at high resolutions. With the aid of fixed reference controls in the form of concrete target surfaces of varying roughness, registration accuracy was better than 1 mm and the mean level of change detection was less than 2.2 mm. Comparisons between the erosion mass estimated from LiDAR scans and from runoff samples suggest that using the LiDAR to monitor soil erosion at the plot scale is feasible, and provides guidance about the level of accuracy one might expect in doing so, and the number of scan positions could be reduced to three while not significantly impacting the volumetric change estimations. The second study investigates the effects of DEM interpolation and interpolation methods on quantification of soil surface roughness. DEMs errors tended to increase after the surfaces evolved to a rougher state and when a coarser resolution was used. Comparisons between the random roughness calculated from LiDAR point cloud directly and from LiDAR-interpolated DEMs showed that DEMs generated from LiDAR underestimated soil surface roughness and are ineffective for tracking changes in soil surface roughness over time and that LiDAR point cloud data must be used instead on the present scale. The third study tracks the spatiotemporal evolution of soil surface roughness on stony plots. Roughness indices including random roughness (RR), fractal dimension and generalized fractal dimension, were calculated from LiDAR point cloud directly. Surface roughness showed an increasing trend as the rainfall simulation proceeded, and the steeper slope resulted in greater surface roughness. Both the increase of surficial exposed rocks and the formations of erosion features, e.g., rills and depressions, contributed to the spatiotemporal variations of soil surface roughness. Results also showed that the fractal dimension was not a good indicator of soil surface roughness, but rather was an index of the form of the surface. Crossover length was a measure of roughness at a scale of a few millimeters, while random roughness was a measure of elevation variation on the scale of the length of the transect measured, and thus encompassed larger morphological features including rills. We also established a new method for analyzing multiple fractals that characterized the heterogeneity of soil surface roughness. These results improve our understanding of the evolution of semiarid stony hillslopes.

Integer optimization (IO) problems arise in research areas and our daily life almost in every aspect. IO formulations and methods can be adopted to make optimal decisions for solution-searching and management with global optimization. Two main areas of applications of IO are studied in the dissertation. Not only the compact formulations for each problem are proposed, but also possible extensions with respect to solution approaches and illustrative case studies are presented. The first area studied in the dissertation is the application of IO to biomass supply chain. The IO is utilized to make various decisions when coordinating the flows in a biomass supply chain network. A typical biomass supply chain contains following five operational components: harvesting and collection, storage, transportation, pretreatment, and conversion. In the process of each component, there are some related classic optimization methods, including linear programming models, integer programming models, stochastic programming models, approximate approaches, etc. A detailed literature review on optimization methods for biomass supply chain is presented in this dissertation. Besides that, the uncertainties and challenges in a biomass supply chain and corresponding optimization methods to handle them are reviewed. Moreover, environmental and social issues arising in a biomass supply chain is also studied in the review. In latter work, a facility location problem is studied since the facility location decision is vital to a supply chain and has long-term influences on both cost and income. Geographic Information System (GIS) analysis is integrated into the optimization method. GIS analysis is used to identify the potential candidates which satisfy all the given criteria. Uncertainty coming from planting plan is also considered in the problem and the problem is formulated as a two-stage stochastic programming model. The second area studied in the dissertation is the application of IO to power systems, especially power dominating set problems. The IO has been applied in almost every aspect of the power system industry, including power system network design problem, unit commitment problem, and sensor placement problem. Phasor Measurement Unit (PMU) placement problem is studied in the following two manners. First, a multistage PMU placement problem is paid attention to due to the high expense of PMUs and budget limitation at each time period. The IO formulation is proposed to maximize the observation of the power system at each stage and guarantee the full observation at the last planning stage. In order to enhance the computational efficiency of the IO formulation, the problem is converted to a multistage network flow problem. Second, uncertain issues are taken into consideration and the probabilistic and reliable connected power dominating set problem is studied. In a power system, the placed PMUs in the power grid and transmission lines may not always function well. The power system health under contingencies is studied. Given pre-specified reliability level for each bus in the power system to be observed and known distribution of random events, the IO formulation is proposed to satisfy the reliability requirement and meet some other problem-specific restrictions at the same time. In the work, the connectivity of the subgraph constructed by PMUs is also guaranteed for better communication of PMUs and the reliability of the connectivity is also studied.