The prospect of using the Internet and other Big Data methods to construct event data promises to transform the field but is stymied by the lack of a coherent strategy for addressing the problem of selection. Past studies have shown that event data have significant selection problems. In terms of conventional standards of representativeness, all event data have some unknown level of selection no matter how many sources are included. We summarize recent studies of news selection and outline a strategy for reducing the risks of possible selection bias, including techniques for generating multisource event inventories, estimating larger populations, and controlling for nonrandomness. These build on a relativistic strategy for addressing event selection and the recognition that no event data set can ever be declared completely free of selection bias.

近年來，許多專家學者廣泛探討偵測稀有疾病的發生率或稱為叢集上的空間或空間對時間的統計方法及模型。這些方法大部分都是處理個別資料或是只能偵測接近圓形的叢集。在這篇論文中，根據Choynowski在1959年所探討的方法，我們進一步提出針對整體資料去偵測非圓形叢集的方法，並且會將此方法與Nagarwalla’s Spatial Scan Statistic做比較。同時，我們會呈現模擬結果中的型一、型二誤差來衡量此方法的可行性。另外，我們也會將此方法實際應用到台灣的癌症死亡資料做探討。 / In recent years, many statistical methods have been proposed for detecting excesses of rare diseases, i.e., clusters, in space or in space-time. Most of these methods deal with case-event or individual-level data and can only detect clusters with shape close to circles. In this study, adapting Choynowski's (1959) idea, a simulation-based approach is proposed to detect non-circular clusters with aggregate or group-level data. The proposed cluster detection method will be used to compare with a frequently used method: Nagarwalla’s Spatial Scan Statistic. Computer simulation is used to illustrate the validity, with respect to Type-I and Type-II errors, of the proposed approach. In addition, the cancer mortality data in Taiwan area are also used as a demonstration of the proposed test.

Joint modeling of a single event time response with a longitudinal covariate dates back to the 1990s. The three basic types of joint modeling formulations are selection models, pattern mixture models and shared parameter models. The shared parameter models are most widely used. One type of a shared parameter model (Joint Model I) utilizes unobserved random effects to jointly model a longitudinal sub-model and a survival sub-model to assess the impact of an internal time-dependent covariate on the time-to-event response. Motivated by the Muscular Dystrophy Surveillance, Tracking and Research Network (MD STARnet), we constructed a new model (Joint Model II), to jointly analyze correlated bivariate time-to-event responses associated with an internal time-dependent covariate in the Frequentist paradigm. This model exhibits two distinctive features: 1) a correlation between bivariate time-to-event responses and 2) a time-dependent internal covariate in both survival models. Developing a model that sufficiently accommodates both characteristics poses a challenge. To address this challenge, in addition to the random variables that account for the association between the time-to-event responses and the internal time-dependent covariate, a Gamma frailty random variable was used to account for the correlation between the two event time outcomes. To estimate the model parameters, we adopted the Expectation-Maximization (EM) algorithm. We built a complete joint likelihood function with respect to both latent variables and observed responses. The Gauss-Hermite quadrature method was employed to approximate the two-dimensional integrals in the E-step of the EM algorithm, and the maximum profile likelihood type of estimation method was implemented in the M-step. The bootstrap method was then applied to estimate the standard errors of the estimated model parameters. Simulation studies were conducted to examine the finite sample performance of the proposed methodology. Finally, the proposed method was applied to MD STARnet data to assess the impact of shortening fractions and steroid use on the onsets of scoliosis and mental health issues.

In the last decade or so, Iran and Israel have found a new domain where to translate and protract their conflictual relationship: the cyberspace. Since the discovery of Stuxnet both countries have embarked on a significant cyber capabilities build-up, in accordance with their resources, and driven as well by mutual threat and perception of threat they pose to each other. Following their own cyber strategies embedded in their respective broader foreign policy agendas, the two foes confronted each other several times in a cyber feud which lasts to present days. Through the use of Event Data methodology, it is examined quantitatively the impact of cyber conflict on the Iranian- Israeli relationship, when cyber capabilities are utilized as a foreign policy tool vis-à-vis each other. The results of the quantitative study show no significant alteration of the conflict-cooperation dynamics between the dyad of interest due to the utilization of cyber capabilities. Furthermore, the qualitative assessment of the cyber feud shows that the balance of power between the two sides is not affected by increasing cyber capabilities, since Israel has the clear upper hand in the cyberspace. Keywords Iran, Israel, Foreign Policy, Cyber Capabilities, Cyber Conflict, Event Data

This thesis presents the results of a research effort investigating the effect of belt usage reporting errors of National Automotive Sampling System-Crash Data System (NASS-CDS) investigators on injury risk estimates. Current estimates of injury risk are developed under the assumption that NASS-CDS investigators are always accurate at determining seat belt usage. The primary purpose of this research is to determine the accuracy of NASS-CDS investigators using event data recorders (EDRs) as the baseline for accuracy, and then recalculating injury risk estimates based on our findings. The analysis of a 107 EDR dataset, from vehicle tests conducted by the National Highway Traffic Safety Administration (NHTSA) and the Insurance Institute for Highway Safety (IIHS), was conducted to determine the accuracy of Chrysler, Ford, GM and Toyota EDRs. This accuracy was examined by both EDR module type and vehicle make. EDR accuracy was determined for crash delta-V, seat belt buckle status, pre-impact speed, airbag deployment status and front seat position. From this analysis we were able to conclude that EDRs were accurate, within 4.5%, when comparing maximum delta-V of EDRs that recorded the entire crash pulse length. We also determined that EDRs were 100% accurate when reporting driver seat belt status for EDRs that completely recorded the event and recorded a status for the driver's seat belt. All GM, Ford and Chrysler EDRs in our database reported a pre-impact velocity less than 6 mph different than the NHTSA and IIHS reported pre-impact velocities. We also found that all but 2 (101 out of 103) of the GM, Ford, and Toyota EDRs correctly reported airbag deployment status. Lastly we were able to conclude that seat position status was useful in determining when a smaller sized occupant was the driver or right front occupant. EDRs reported seat position of 5% Hybrid III females as "forward" in every case that seat position was recorded for this smaller occupant size. Based on the analysis of seat belt status accuracy, a comparison of NASS-CDS investigator driver seat belt status and EDR driver seat belt status was conducted to determine the accuracy of the NASS-CDS investigators. This same comparison was conducted on reports of driver seat belt status provided by police. We found that NASS-CDS investigators had an overall error of 9.5% when determining driver seat belt status. When the EDR stated that the driver was unbuckled, investigators incorrectly coded buckled in of 29.5% of the cases. When the EDR stated that the driver was buckled, NASS-CDS error was only 1.2%. Police officers were less accurate than NASS-CDS investigators, with an overall error of 21.7%. When the EDR stated that the driver was buckled, police had an error of 2.4%. When the EDR stated that the driver's belt was unbuckled, police had an error of around 69%. In 2008, NASS-CDS investigators reported that drivers had an overall belt usage rate during accidents of 82%. After correcting for the errors we discovered, we estimate that the driver belt buckle status during a crash is around 72.6%. Injury risk estimates and odd ratio point estimates were then calculated for NASS-CDS investigator and EDR buckled versus unbuckled cases. The cases included only frontal collisions in which there was no rollover event or fire. Injury was defined as AIS 2+. The risk ratios and point estimates were then compared between investigators and EDRs. We found that injury risk for unbelted drivers may be over estimated by NASS-CDS investigators. The unbuckled to buckled risk ratio for EDRs was 8%-12% lower than the risk ratio calculated for NASS-CDS investigators. / Master of Science

<p>Event data provide high-resolution and high-volume information about political events. From COPDAB to KEDS, GDELT, ICEWS, and PHOENIX, event datasets and the frameworks that produce them have supported a variety of research efforts across fields and including political science. While these datasets are machine-coded from vast amounts of raw text input, they nonetheless require substantial human effort to produce and update sets of required dictionaries. I introduce a novel method for generating large dictionaries appropriate for event-coding given only a small sample dictionary. This technique leverages recent advances in natural language processing and deep learning to greatly reduce the researcher-hours required to go from defining a new domain-of-interest to producing structured event data that describes that domain. An application to cybersecurity is described and both the generated dictionaries and resultant event data are examined. The cybersecurity event data are also examined in relation to existing datasets in related domains.</p> / Dissertation

This thesis considers a variety of statistical issues related to the design and analysis of clinical trials involving multiple lifetime events. The use of composite endpoints, multivariate survival methods with dependent censoring, and recurrent events with dependent termination are considered. Much of this work is based on problems arising in oncology research. Composite endpoints are routinely adopted in multi-centre randomized trials designed to evaluate the effect of experimental interventions in cardiovascular disease, diabetes, and cancer. Despite their widespread use, relatively little attention has been paid to the statistical properties of estimators of treatment effect based on composite endpoints. In Chapter 2 we consider this issue in the context of multivariate models for time to event data in which copula functions link marginal distributions with a proportional hazards structure. We then examine the asymptotic and empirical properties of the estimator of treatment effect arising from a Cox regression model for the time to the first event. We point out that even when the treatment effect is the same for the component events, the limiting value of the estimator based on the composite endpoint is usually inconsistent for this common value. The limiting value is determined by the degree of association between the events, the stochastic ordering of events, and the censoring distribution. Within the framework adopted, marginal methods for the analysis of multivariate failure time data yield consistent estimators of treatment effect and are therefore preferred. We illustrate the methods by application to a recent asthma study. While there is considerable potential for more powerful tests of treatment effect when marginal methods are used, it is possible that problems related to dependent censoring can arise. This happens when the occurrence of one type of event increases the risk of withdrawal from a study and hence alters the probability of observing events of other types. The purpose of Chapter 3 is to formulate a model which reflects this type of mechanism, to evaluate the effect on the asymptotic and finite sample properties of marginal estimates, and to examine the performance of estimators obtained using flexible inverse probability weighted marginal estimating equations. Data from a motivating study are used for illustration. Clinical trials are often designed to assess the effect of therapeutic interventions on occurrence of recurrent events in the presence of a dependent terminal event such as death. Statistical methods based on multistate analysis have considerable appeal in this setting since they can incorporate changes in risk with each event occurrence, a dependence between the recurrent event and the terminal event and event-dependent censoring. To date, however, there has been limited methodology for the design of trials involving recurrent and terminal events, and we addresses this in Chapter 4. Based on the asymptotic distribution of regression coefficients from a multiplicative intensity Markov regression model, we derive sample size formulae to address power requirements for both the recurrent and terminal event processes. Superiority and non-inferiority trial designs are dealt with. Simulation studies confirm that the designs satisfy the nominal power requirements in both settings, and an application to a trial evaluating the effect of a bisphosphonate on skeletal complications is given for illustration.

Multiple imputation (MI) is a commonly used approach to impute missing data. This thesis studies missing covariates in recurrent event data, and discusses ways to include the survival outcomes in the imputation model. Some MI methods under consideration are the event indicator D combined with, respectively, the right-censored event times T, the logarithm of T and the cumulative baseline hazard H0(T). After imputation, we can then proceed to the complete data analysis. The Cox proportional hazards (PH) model and the PWP model are chosen as the analysis models, and the coefficient estimates are of substantive interest. A Monte Carlo simulation study is conducted to compare different MI methods, the relative bias and mean square error will be used in the evaluation process. Furthermore, an empirical study based on cardiovascular disease event data which contains missing values will be conducted. Overall, the results show that MI based on the Nelson-Aalen estimate of H0(T) is preferred in most circumstances.

This thesis considers a variety of statistical issues related to the design and analysis of clinical trials involving multiple lifetime events. The use of composite endpoints, multivariate survival methods with dependent censoring, and recurrent events with dependent termination are considered. Much of this work is based on problems arising in oncology research. Composite endpoints are routinely adopted in multi-centre randomized trials designed to evaluate the effect of experimental interventions in cardiovascular disease, diabetes, and cancer. Despite their widespread use, relatively little attention has been paid to the statistical properties of estimators of treatment effect based on composite endpoints. In Chapter 2 we consider this issue in the context of multivariate models for time to event data in which copula functions link marginal distributions with a proportional hazards structure. We then examine the asymptotic and empirical properties of the estimator of treatment effect arising from a Cox regression model for the time to the first event. We point out that even when the treatment effect is the same for the component events, the limiting value of the estimator based on the composite endpoint is usually inconsistent for this common value. The limiting value is determined by the degree of association between the events, the stochastic ordering of events, and the censoring distribution. Within the framework adopted, marginal methods for the analysis of multivariate failure time data yield consistent estimators of treatment effect and are therefore preferred. We illustrate the methods by application to a recent asthma study. While there is considerable potential for more powerful tests of treatment effect when marginal methods are used, it is possible that problems related to dependent censoring can arise. This happens when the occurrence of one type of event increases the risk of withdrawal from a study and hence alters the probability of observing events of other types. The purpose of Chapter 3 is to formulate a model which reflects this type of mechanism, to evaluate the effect on the asymptotic and finite sample properties of marginal estimates, and to examine the performance of estimators obtained using flexible inverse probability weighted marginal estimating equations. Data from a motivating study are used for illustration. Clinical trials are often designed to assess the effect of therapeutic interventions on occurrence of recurrent events in the presence of a dependent terminal event such as death. Statistical methods based on multistate analysis have considerable appeal in this setting since they can incorporate changes in risk with each event occurrence, a dependence between the recurrent event and the terminal event and event-dependent censoring. To date, however, there has been limited methodology for the design of trials involving recurrent and terminal events, and we addresses this in Chapter 4. Based on the asymptotic distribution of regression coefficients from a multiplicative intensity Markov regression model, we derive sample size formulae to address power requirements for both the recurrent and terminal event processes. Superiority and non-inferiority trial designs are dealt with. Simulation studies confirm that the designs satisfy the nominal power requirements in both settings, and an application to a trial evaluating the effect of a bisphosphonate on skeletal complications is given for illustration.

This dissertation presents a unique approach to human behavior analysis based on expert guidance and intervention through interactive construction and modification of behavior models. Our focus is to introduce the research area of behavior analysis, the challenges faced by this field, current approaches available, and present a new analysis approach: Interactive Relevance Search and Modeling (IRSM). More intelligent ways of conducting data analysis have been explored in recent years. Ma- chine learning and data mining systems that utilize pattern classification and discovery in non-textual data promise to bring new generations of powerful "crawlers" for knowledge discovery, e.g., face detection and crowd surveillance. Many aspects of data can be captured by such systems, e.g., temporal information, extractable visual information - color, contrast, shape, etc. However, these captured aspects may not uncover all salient information in the data or provide adequate models/patterns of phenomena of interest. This is a challenging problem for social scientists who are trying to identify high-level, conceptual patterns of human behavior from observational data (e.g., media streams). The presented research addresses how social scientists may derive patterns of human behavior captured in media streams. Currently, media streams are being segmented into sequences of events describing the actions captured in the streams, such as the interactions among humans. This segmentation creates a challenging data space to search characterized by non- numerical, temporal, descriptive data, e.g., Person A walks up to Person B at time T. This dissertation will present an approach that allows one to interactively search, identify, and discover temporal behavior patterns within such a data space. Therefore, this research addresses supporting exploration and discovery in behavior analysis through a formalized method of assisted exploration. The model evolution presented sup- ports the refining of the observer\'s behavior models into representations of their understanding. The benefit of the new approach is shown through experimentation on its identification accuracy and working with fellow researchers to verify the approach\'s legitimacy in analysis of their data. / Ph. D.