Climate change is impacting the intensity, duration, and frequency of climatic events. With climate change comes a multitude of adverse conditions, including extreme heat events, changes in disease patterns, and increased likelihood and frequency of natural disasters, including in places previously not exposed to such conditions. Human health has foundations in the environment; therefore, these adverse climatic conditions are directly linked to human health. Rural communities in Appalachia are likely to experience negative consequences of climate change more severely due to unique geomorphology and sociopolitical realities of the region. Non-governmental organizations (NGOs) throughout the Appalachian region are currently working to build resilience and prepare for potential adverse effects from climate change. To aid in this process, projections of future climate scenarios are needed to understand possible situations and adequately prepare. In partnership with Ohio University and West Virginia University, this study aims to characterize potential future climatic scenarios from publicly-available global climate models (GCMs) and prepare information to share with Appalachian communities. Climate model information for this analysis was obtained from NASA's Coupled Model Intercomparison Project (CMIP6). All code for data processing and analysis was prepared using the open-source R programming language to support reproducibility. To confirm that models can accurately simulate Appalachian climatic conditions, CMIP6 hindcast simulations for precipitation and maximum temperature were compared to observed weather records from NOAA. Climate models over and underestimated average precipitation values depending on location, while models consistently underestimated extreme precipitation values, simulated by total five-day precipitation. For temperature, climate models consistently underestimated average and extreme high temperature indicators. For Appalachian region projections, three towns of interest (one for each state involved in the study: Virginia, West Virginia, and Ohio) were selected based on current community resilience efforts. In these locations, mid-century (2040 – 2064) and end-of-century (2075 – 2099) projections for precipitation and temperature were summarized under a low emissions scenario and a high emissions scenario. Increases in precipitation and temperature were observed under average and extreme scenarios; these increases were noticeably more extreme under higher emissions scenarios. These trends are consistent with other studies and climate science consensus. When compared to hindcast values, observed average precipitation values were overestimated and underestimated, while observed extreme precipitation indices, average temperatures, and heat wave indices were underestimated by GCMs. Context with observed data is important to understanding model accuracy for the Appalachian region. GCMs are a useful tool to project potential future climate scenarios at specific locations in the Appalachian region, though model data is best used to communicate general trends rather than as inputs for other physical models. / Master of Science / Climate change is driving previously unseen changes in many aspects of the environment. Among these aspects, and of particular concern, are increased precipitation and increased high temperatures, which have direct negative outcomes on human health. Climate change can impact human health in a variety of ways, such as increasing instances of heat-related illnesses like heatstroke, changing insect-carried diseases patterns (i.e. Lyme disease, malaria), worsening preexisting conditions like asthma, and increasing the likelihood of natural disasters like flooding. Climate change also impacts mental health, especially increasing instances of anxiety and post-traumatic stress disorder from disasters. Rural communities like Appalachia are more likely to experience severe negative outcomes due to lack of resources, remote location, and economies historically based on resource extraction. Appalachia specifically also faces unique challenges with flooding, as many towns are situated in valleys with streams or rivers running through the center of town. To address and prepare for possible climate change outcomes, community-based planning is required to build resiliency. Throughout many areas, but specifically in Appalachia, many community-based organizations are already working to strengthen their communities by providing stable housing, addressing flooding, and preparing emergency response teams. To aid in these efforts, information about potential future climate is beneficial to these organizations to understand and prepare for potential conditions. This study aims to use publicly-available climate models to generate information about possible future climate conditions to be shared with community organizations. Additionally, this project's datasets and procedures are publicly available, so this analysis can be performed by communities anywhere in the world given they have adequate computing power. To check that models are a good indicator of previous climate conditions, and therefore would be useful for future projections, historic projected climate model outputs were compared to observed weather data. After confirming that the models used were fairly consistent with observed data, projected values for midcentury (2040 – 2064) and end-of-century (2075 – 2099) were gathered for Appalachian towns with interested community organizations. Projected values show increases in high temperatures and precipitation throughout the Appalachian region, including in short-term event scenarios, which is consistent with other climate science. Higher emissions scenarios result in greater increases in average and extreme temperature and precipitation values. Climate models can be a useful tool in understanding potential general climatic trends for a specific location and can support climate science communication.

The current study examined the nature and extent of endorsement of PMIEs, the nature and severity of MI symptoms related to endorsement of a PMIE, and the relations between extent of endorsement of PMIEs and MI symptoms. We hypothesized that (1) PMIEs perpetrated by others would be endorsed to a greater extent than PMIEs perpetrated by oneself; (2) medical professionals who endorsed a PMIE would report significantly greater severity on all MI symptoms compared to medical professionals who did not endorse a PMIE; (3) experiencing PMIEs (perpetrated by oneself and/or others) to a greater extent would predict higher levels of MI symptom severity, and MI symptom severity would specifically be most strongly predicted by PMIEs perpetrated by oneself. Hypotheses were examined using t-tests, Pearson's r correlations, and multiple multivariate regression analyses. First, the current study found that PMIEs perpetrated by others were endorsed to a greater extent than those perpetrated by themselves; second, greater exposure to PMIEs was associated with significantly greater severity of 10 of the 14 outcomes. Third, PMIEs perpetrated by oneself predicted more MI symptomatology than PMIEs perpetrated by others, indicating that while PMIEs perpetrated by others are more common, PMIEs perpetrated by oneself are more strongly associated with MI outcomes. This study highlights the widespread and harmful impact of PMIEs among medical professionals.

The design and development of technology applications has to deal with many variables. Reference is obviously made to established hardware and software support, particularly with regard to the choice of appropriate operating systems, development model, environment and programming language. With the growth of networked and web-exposed systems, we are increasingly dealing with IoT (Internet-of-Things) systems: complex applications consisting of a network of often heterogeneous elements to be managed like an orchestra, using existing elements and creating new ones. Among the many fields affected by this phenomenon, two in particular are considered here: industry and medical, key sectors of modern society. Given the inherently parallel nature of such networks and the fact that it is commonly necessary to manage them via the Web, the most prevalent de facto model employs an architecture relying on a paradigm based on data flows, representing the entire system as a kind of assembly line in which each entity acquires input data and returns an output in a perfectly asynchronous manner. This thesis highlights some notable limitations of this approach and proposes an evolution that resolves some key issues. This has been done not only on a purely theoretical level, but with actual implementations currently operational and thus demonstrated in the field. Rather than proposing an abstract formalisation of a new solution, the basic principles of a whole new architecture are presented here instead, going into more detail on some key features and with experimental and practical feedback implemented as a full blown software platform. A first contribution is the definition of the principles of a new programming architecture, disseminated with some published articles and a speech in an international congress. A second contribution concerns a lightweight data synchronisation strategy, which is particularly useful for components that need to continue working during offline periods. A third contribution concerns a method of storing a symmetric encryption key combined with a peculiar retrieval and verification technique: this has resulted in an international patent, already registered. A fourth contribution concerns a new data classification model, which is particularly effective for processing information asynchronously. Issues related to possible integrations with artificial intelligence systems have also been addressed, for which a number of papers are being written, introduced by a presentation that has just been published.

The importance of traffic safety has prompted considerable research on predicting driver risk and evaluating the impact of risk factors. Driver risk modeling is challenging due to the rarity of motor vehicle crashes and heterogeneity in individual driver risk. Statistical modeling and analysis of such driver data are often associated with Big Data, considerable noise, and lacking informative predictors. This dissertation aims to develop several systematic techniques for traffic safety modeling, including finite sample bias correction, decision-adjusted modeling, and effective risk factor construction. Poisson and negative binomial regression models are primary statistical analysis tools for traffic safety evaluation. The regression parameter estimation could suffer from the finite sample bias when the event frequency (e.g., the total number of crashes) is low, which is commonly observed in safety research. Through comprehensive simulation and two case studies, it is found that bias adjustment can provide more accurate estimation when evaluating the impacts of crash risk factors. I also propose a decision-adjusted approach to construct an optimal kinematic-based driver risk prediction model. Decision-adjusted modeling fills the gap between conventional modeling methods and the decision-making perspective, i.e., on how the estimated model will be used. The key of the proposed method is to enable a decision-oriented objective function to properly adjust model estimation by selecting the optimal threshold for kinematic signatures and other model parameters. The decision-adjusted driver-risk prediction framework can outperform a general model selection rule such as the area under the curve (AUC), especially when predicting a small percentage of high-risk drivers. For the third part, I develop a Multi-stratum Iterative Central Composite Design (miCCD) approach to effectively search for the optimal solution of any "black box" function in high dimensional space. Here the "black box" means that the specific formulation of the objective function is unknown or is complicated. The miCCD approach has two major parts: a multi-start scheme and local optimization. The multi-start scheme finds multiple adequate points to start with using space-filling designs (e.g. Latin hypercube sampling). For each adequate starting point, iterative CCD converges to the local optimum. The miCCD is able to determine the optimal threshold of the kinematic signature as a function of the driving speed. / Doctor of Philosophy / When riding in a vehicle, it is common to have personal judgement about whether the driver is safe or risky. The drivers’ behavior may affect your opinion, for example, you may think a driver who frequently hard brakes during one trip is a risky driver, or perhaps a driver who almost took a turn too tightly may be deemed unsafe, but you do not know how much riskier these drivers are compared to an experienced driver. The goal of this dissertation is to show that it is possible to quantify driver risk using data and statistical methods. Risk quantification is not an easy task as crashes are rare and random events. The wildest driver may have no crashes involved in his/her driving history. The rareness and randomness of crash occurrence pose great challenges for driver risk modeling. The second chapter of this dissertation deals with the rare-event issue and provides more accurate estimation. Hard braking, rapid starts, and sharp turns are signs of risky driving behavior. How often these signals occur in a driver’s day-to-day driving reflects their driving habits, which is helpful in modeling driver risk. What magnitude of deceleration would be counted as a hard brake? How hard of a corner would be useful in predicting high-risk drivers? The third and fourth chapter of this dissertation attempt to find the optimal threshold and quantify how much these signals contribute to the assessment of the driver risk. In Chapter 3, I propose to choose the threshold based on the specific application scenario. In Chapter 4, I consider the threshold under different speed limit conditions. The modeling and results of this dissertation will be beneficial for driver fleet safety management, insurance services, and driver education programs.

Do people become more optimistic about future adjustment to negative life events after rumination? Past research using a "top of the head" paradigm indicates that people estimate they would adjust more poorly for severe events and better for mild negative events than their peers. Selective focus (i.e., differential accessibility of information about assets and liabilities for coping) has been provided as an explanation for this effect, which is counter to research on "optimistic bias". Martin and Tesser's (1989) rumination model was applied to beliefs about one's comparative adjustment to negative life events. One hundred twenty undergraduate subjects were asked to imagine experiencing a Severe (HIV+) or Mild (Herpes) negative event at some future time, then to designate items on a reaction time task as either an Asset or Liability in coping with the event. The reaction time task and subsequent comparative adjustment ratings were made either immediately, after a delay that allowed for rumination, or after a delay without an opportunity for rumination. A thought-listing analysis of the audiotaped ruminations revealed that, as predicted, subjects became more optimistic over time. They initially discussed liabilities in coping with the Severe event but gradually considered assets. Comparative adjustment ratings for the Severe event were not significantly different than for the Mild event, even in the Rumination Absent condition. It was suggested that temporarily making assets for coping accessible through the reaction time task had the same effect on comparative adjustment ratings as did problem-solving through rumination. The reaction time data provided convergent evidence regarding selective focus and complimented a thought-listing paradigm used in previous studies. / Master of Science

This dissertation demonstrates how patients' mental illness treatment careers depend on the change and/or stability among differing levels of social structure. Theorists of the mental illness career tend to ignore the role that higher levels of social structural change have on individuals' mental illness career. Researchers using an organizational perspective tend to focus on the organizational environment but ignore the treatment process from the individual's point of view. Both perspectives neglect what the nation-state's broader socio-political and economic circumstances could imply for people seeking treatment for mental disorders. Organizational theory and theories of the mental illness career are independent theoretical streams that remain separate. This dissertation connects these independent theoretical streams by developing a unifying theoretical framework based on historical analysis. This historical analysis covers three phases of treatment beginning at the end of World War II to the present. This framework identifies mechanisms through which changes in larger levels of social structure can change the experience and career of mental patients. This new perspective challenges current conceptions of the mental illness career as static by accounting for the various levels of social structure that play a part in the mental illness treatment career. Taken together, the inclusion of differing levels of social structure and the subsequent reciprocal relationship between these levels of analysis produce a narrative that explains why and how stability and change within the mental health sector shape the mental illness treatment career. / Ph. D.

Corporate crisis events such as cyber attacks, executive scandals, facility accidents, fraud, and product recalls can damage customer trust and firm reputation severely, which may lead to tremendous loss in sales and firm equity value. My research aims to integrate information available on the market to assist firms in tackling crisis events, and to provide insight for better decision making. We first study the impact of crisis events on firm performance. We build a hybrid deep learning model that utilizes information from financial news, social media, and historical stock prices to predict firm stock performance during firm crisis events. We develop new methodologies that can extract, select, and represent useful features from textual data. Our hybrid deep learning model achieves 68.8% prediction accuracy for firm stock movements. Furthermore, we explore the underlying mechanisms behind how stakeholders adopt and propagate event information on social media, as well as how this would impact firm stock movements during such events. We adopt an extended epidemiology model, SEIZ, to simulate the information propagation on social media during a crisis. The SEIZ model classifies people into four states (susceptible, exposed, infected, and skeptical). By modeling the propagation of firm-initiated information and user-initiated information on Twitter, we simulate the dynamic process of Twitter stakeholders transforming from one state to another. Based on the modeling results, we quantitatively measure how stakeholders adopt firm crisis information on Twitter over time. We then empirically evaluate the impact of different information adoption processes on firm stock performance. We observe that investors often react very positively when a higher portion of stakeholders adopt the firm-initiated information on Twitter, and negatively when a higher portion of stakeholders adopt user-initiated information. Additionally, we try to identify features that can indicate the firm stock movement during corporate events. We adopt Layer-wised Relevance Propagation (LRP) to extract language features that can be the predictive variables for stock surge and stock plunge. Based on our trained hybrid deep learning model, we generate relevance scores for language features in news titles and tweets, which can indicate the amount of contributions these features made to the final predictions of stock surge and plunge. / Doctor of Philosophy / Corporate crisis events such as cyber attacks, executive scandals, facility accidents, fraud, and product recalls can damage customer trust and firm reputation severely, which may lead to tremendous loss in sales and firm equity value. My research aims to integrate information available on the market to assist firms in tackling crisis events and providing insight for better decision making. We first study the impact of crisis events on firm performance. We investigate five types of crisis events for SandP 500 companies, with 14,982 related news titles and 4.3 million relevant tweets. We build an event-driven hybrid deep learning model that utilizes information from financial news, social media, and historical stock prices to predict firm stock performance during firm crisis events. Furthermore, we explore how stakeholders adopt and propagate event information on social media, as well as how this would impact firm stock movements during the events. Social media has become an increasingly important channel for corporate crisis management. However, little is known on how crisis information propagates on social media. We observe that investors often react very positively when a higher portion of stakeholders adopt the firm-initiated information on Twitter, and negatively when a higher portion of stakeholders adopt user-initiated information. In addition, we find that the language used in the crisis news and social media discussions can have surprising predictive power on the firm stock. Thus, we develop a methodology to identify the importance of text features associated with firm performance during crisis events, such as predictive words or phrases.

<p dir="ltr">Tornadoes have caused billions of dollars in damage and are one of the leading causes of weather-related deaths in the United States each year. Recent studies have suggested spatial shifts in tornado activity, though the reason is unclear. Extratropical cyclones (ETCs), which are strongly associated with the jet stream, are known to produce an environment favorable for tornadoes in their warm sector. However, little recent research has been done on the spatiotemporal relationship between tornadoes and ETCs, so there is a poor understanding of whether or not the changes in tornado activity are affected by ETC patterns. ERA5 reanalysis, ETC tracking, and historical tornado data from 1980-2022 are used to analyze the climatology of tornadoes relative to ETCs. We found that 73% of F/EF1+ tornadoes occurred within 2000km of an ETC and are likely associated with the ETC. Most of those tornadoes occurred near the median position around 465km away from and to the southeast of the ETC center. Of those tornadoes, 68% occurred in large outbreaks of 6 or more tornadoes, where most tornadoes formed closer to and to the southeast of an ETC track as compared to small outbreak and isolated tornadoes. The spatial and relative distributions were similar across all intensity levels, though stronger tornadoes tended to have more tornadoes directly to the southeast of an ETC. Seasonal variances in tornadoes strongly corresponded with seasonal changes in the jet stream. Summer tornadoes occurred in northern portions of the US when the jet stream shifts poleward. The jet stream and ETCs are also weakest in the summer, and the weakest association was found in summer tornadoes based on their distribution relative to ETCs being more uniform towards the northeast and north-southeast directions. Winter tornadoes occurred in more southern portions of the US when the jet stream shifts equatorward, and they had a stronger association with most of the tornadoes occurring to the southeast and closer to the ETC center, aligning with a strong ETC and jet stream in the winter. Finally, tornadoes and ETCs had strong spatial covariance and showed similar linear trends, including a similar rate of change in the eastward shift, providing strong evidence that a shift in ETCs may be driving the shift in tornadoes. Furthering our understanding of the relationship between tornadoes and ETCs will help to better predict how tornadoes will change in the future based on changes in ETCs.</p>

Transarterial chemoembolization (TACE) and transarterial radioembolization (TARE) are recommended to treat patients with early or intermediate hepatocellular carcinoma (HCC). The liver maximum capacity test (LiMAx) has been supposed to predict the risk of post-interventional liver failure. We investigated the correlation of LiMAx with short term survival as primary endpoint and the occurrence of adverse events after therapy as secondary endpoint. Our study cohort prospectively included 69 patients receiving TACE (n = 57) or TARE (n = 12). LiMAx test and serological analyses were performed on the day before and 4 weeks after treatment. Hepatic and extrahepatic complications were monitored for 4 weeks. The LiMAx results were not associated with altered liver function and the occurrence of adverse events. The survival rates of patients with BCLC A with LiMAx ≤ 150 μg/kg/h were lower after 30 days (75.0 ± 15.3% vs. 100%, p = 0.011), 90 days (62.5 ± 17.7% vs. 95.8 ± 4.1%, p = 0.011) and 180 days (50.0 ± 17.7% vs. 95.8 ± 4.1%, p = 0.001) compared to those with higher LiMAx levels. The LiMAx test is not suitable to predict liver function abnormalities or the occurrence of complications 4 weeks after therapy but enables the identification of patients with early stage HCC and reduced short-term survival after treatment.

Doctor of Philosophy / With the fast development of artificial intelligence (AI) technology, the reliability of AI needs to be investigated for confidently using AI products in our daily lives. This dissertation includes three projects introducing the statistical models and model estimation methods that can be used in the reliability analysis of AI systems. The first project analyzes the recurrent events data from autonomous vehicles (AVs). A nonparametric model is proposed to study the reliability of AI systems in AVs, and a statistical framework is introduced to evaluate the adequacy of using traditional parametric models in the analysis. The proposed model and framework are then applied to analyze AV data from four manufacturers that participated in an AV driving testing program overseen by the California Department of Motor Vehicles. The second project develops a survival model to investigate the failure times of graphics processing units (GPUs) used in supercomputers. The model considers several covariates, the spatial correlation, and the correlation among multiple types of failures. In addition, unique spatial correlation functions and a special distance function are introduced to quantify the spatial correlation inside supercomputers. The model is applied to explore the GPU failure times in the Titan supercomputer. The third project proposes a new Markov chain Monte Carlo sampler that can be used in the estimation and inference of spatial survival models. The sampler can generate a reasonable amount of samples within a shorter computing time compared with existing popular samplers. Important factors that can influence the performance of the proposed sampler are explored, and the sampler is used to analyze the Titan GPU failures to illustrate its usefulness in solving real-world problems.