University of Maryland, Baltimore County
Not a member yet
17643 research outputs found
Sort by
EXECUTING SOCIAL INEQUALITY: PREDICTIVE FACTORS IN THE APPLICATION OF CAPITAL PUNISHMENT
No full textThe adoption and implementation of the death penalty varies greatly by state. This dissertation will attempt to explain these differences in terms of the characteristics identified as salient in the empirical and theoretical literature at the state level. Some of the questions addressed include: Are the differences among the states the result of Social Conditions, such as demographics and inequality? Do the political tendencies in political affiliations of the state legislatures explain why some states adopt capital punishment and execute and others do not? Do crime rates, in terms of homicide rates and hate crime rates, have an effect on the implementation of death penalty statutes? Using social control and social dominance theories, independent variables were selected to test these theories. Due to the fact that executions occur in only a few states, zero-inflated negative binomial regression models will be used to test bivariate and multivariate associations (Mwalili, Lesaffre, and Declerck 2007). Percentage of the population that African American, inequality rates, homicide rates, educational attainment, percentage of the population that is evangelical, hate crime rates, exonerations and percentage of the state legislature that is Republican are the independent variables analyzed and number of executions is the dependent variables analyzed. In addition, the percentage of the population that is African American was combined with the percentage of the population that is Hispanic was tested as well. All models controlled for region and year, while state population per million was used as an exposure variable, and the procedures were repeated in models lagged by 5 years. Results indicate that most of the independent variables are associated with the adoption of the death penalty in the bivariate modes (both non-lagged and lagged models). However, in the multivariate models, only hate crime rates and evangelicals were significant in explaining the differences in executions in the non-lagged models when region was removed. Evangelicals, hate crimes, and the combined race variable (% Black + % Hispanic) were significant in predicting states with the death penalty in the lagged models, while homicide rates were only significant in the first lagged model in predicting the adoption of capital punishment. These results support part of the framework used in this research, which includes features of sociological theories for understanding state level variation in the implementation of the death penalty. These results also bolster the argument that state policies are not merely reactions to murder rates, but are influenced by other social factors
Optimization and characterization of self-assembled monolayer multilayer surface-enhanced Raman scattering substrate for immuno-nanosensors
No full textSurface-enhanced Raman scattering (SERS) is a powerful analytical tool and the recent expansion of substrates for SERS measurement broadens its field of applications, including SERS-based immuno-nanosensing. This dissertation describes the optimization and characterization of self-assembled monolayer (SAM) multilayer SERS substrates for improving performance of different types of substrates. SERS substrates derived from metal film on nanostructures were modified with multiple metal films interspaced with SAM dielectric spacers to achieve multilayered SERS substrates. The fundamental concept of this substrate geometry exploited the cumulative effect of the multiple electromagnetic fields and various properties of SAMs to optimize SERS enhancement of multilayer SERS substrates to about 20-fold compared to conventional substrates. SAMs of less bulky terminal groups of alkylthiols with relatively small dielectric constant, strong attractive interaction, and good electron donating ability enhanced multilayer SERS. Compared to SFON SERS substrates, alkyl-terminated SAM substrates exhibited about 7-fold SERS enhancement. SAM formation conditions that enhanced strong intermolecular interactions improved SERS enhancement. Thus, -COOH-terminated SAM multilayer substrates fabricated in acidic conditions exhibited SERS enhancement up to 10-fold compared to similar substrates fabricated in basic conditions. It was observed that the distance of separation of the adjacent metal films (<2 nm) was important for optimum interaction of EM fields generated on these metal films. SAM multilayer SERS substrates formed from 7 or 8 carbon atoms of alkyl-terminated alkylthiols with separation distances of 1.2 and 1.4 nm respectively exhibited about 14 to 16-fold SERS enhancement relative to SFON. A systematic increase of the separation distance between adjacent metal films beyond 2 nm led to a decrease in the SERS enhancement, indicating decreasing interactive effect between the adjacent EM fields. In general, there was an increase in the SERS enhancement with increase in the number of SERS-active layers. Correspondingly, the relative standard deviation decreased to about 5% for multilayered substrates compared to 14% for conventional single layered substrates, thereby improving multilayer SERS reproducibility while increasing the SERS enhancement. Overall, SERS enhancement of conventional SERS substrates can be optimized using multiple SERS-active surfaces well separated with appropriate dielectric spacers and optimum separation distance
The Hispanic-Asian achievement gap in elementary school
No full textThere is little research of Hispanic and Asian children's educational outcomes; in particular, the achievement gap between these two racial/ethnic groups has not been fully explored. The objective of this investigation is to analyze the Hispanic-Asian achievement gap in elementary school using the ECLS-K, a longitudinal nationally representative study. I provide detailed descriptive analysis of the trajectory of the gap from kindergarten to fifth grade. Analysis is conducted using three approaches. First, I estimated the mean achievement gap in elementary school and the relative contribution of family, school and teacher factors to the gap. Second, I decomposed the Hispanic-Asian gap into explained and unexplained parts due to observed differences. Finally, I explored whether the achievement gap is persistent across the conditional distribution of test scores. Reading and math scores are the outcome variables. It is found that Hispanic students enter kindergarten with math and reading skills significantly lower as compared to Asians. In kindergarten the unadjusted reading gap starts at .668 standard deviations below the mean, steadily narrows until third grade but flattens until fifth grade. Math unadjusted gap is larger and presents a different trajectory over time; it starts at .816 standard deviations below the mean, narrows in first grade and broadens again in fifth grade. Overall, I find that the main predictors of the Hispanic-Asian achievement gap are family socioeconomic factors. Teacher and school factors have a marginal predictive power in explaining the gap. Decomposition analysis indicate that differences in observables between Hispanics and Asians children in family, school and teacher factors explain between 22% and 49% of the reading achievement gap in kindergarten. By fifth grade, these factors explain between 34% and 62% of the observed achievement gap; I find similar a pattern for math. Quantile regression estimates shows that the achievement gap is persistent across the entire distribution of test scores. However I find a shift in the pattern of the achievement gap as children progress in the school years. I finalize my analysis by concluding that out-school-factors are the major drivers of the Hispanic-Asian achievement gap and recommend policy interventions aimed at enhance educational attainment amongst socioeconomic disadvantaged Hispanic children
Coronary Heart Disease Prediction:A Data Mining Approach
No full textData mining is a field of computer science that combines statistical analysis and machine learning to detect hard-to-discern patterns from large amounts of data. It employs different algorithms to learn different patterns from training or experience and apply it to classify, predict or identify patterns. The healthcare environment is very information rich. There is a wealth of clinical data available within the healthcare systems. Also due to recent advancement of genomic research vast amount of genetic data are also available. Effective analysis tools are needed to discover hidden relationships and trends in these data. These tools are necessary to correctly diagnose people at risk of disease based on the derived knowledge from the data. We used data mining techniques to evaluate the interaction between traditional risk factors and gene variants such as Single Nucleotide Polymorphisms (SNPs) towards Coronary Heart Disease (CHD) susceptibility in a prospective study of older population aged 65 and older. In our thesis we asked two questions whether we can predict CHD at birth or adding genetic information to traditional risk factors predict CHD better than traditional risk factors alone.We analyzed two popular machine learning algorithms to determine the most efficient method on given domain. We also applied a clustering method to identify different subgroups present in the selected datasets and determine the effect of genetic data on clustering. This study demonstrates the concept of using multiple SNPs as independent risk factors and indicates that it can improve prediction of incident CHD
Joint Inference for Extracting Soft Biometric Text Descriptors from Patient Triage Images
No full textDisaster events can result in mass casualties and missing persons, giving rise to a need to provide information about victims to the public. This can be achieved by digitally documenting information available at emergency medical care centers in the form of pictures. The images and other identifying information, such as fingerprints, cannot be broadcast due to privacy concerns, leading to a need to extract appearance--related non-unique features from this data to facilitate locating missing persons. Using humans and machines to compare images is not feasible due to the scale of the situation and the nature (presence of blood and debris) of the images. Extracting a soft biometric text descriptor (text labels describing different soft biometric features) makes it possible to organize information about individuals from these images in a searchable format without revealing the person's identity. The main aim of this thesis is to extract soft biometric features from person images to label appearance--related information and make it available as a text descriptor. We begin by presenting soft biometric feature detectors for patient images that include an ensemble--based face detection algorithm, template--based eye detection, and eyeglasses, hair color, and skin color detection. We also present a facial hair detector that uses a combination of face and hair information. The feature detection results indicate a need to combine and exploit feature relationships for better performance. We propose a novel probabilistic graphical model that consists of different feature detectors and exploits relationships between these features using a message--passing inference algorithm to build a coherent text descriptor. Further, to understand the utility and the nature of the text descriptors, we present a study based on human descriptions that aims at extracting order and structure information about the features. We evaluate the performance of individual feature detectors for standard and triage images and establish the challenges posed by the dataset. Further, our text analysis shows extreme variability in human descriptions. However, we succeed in extracting some insights about the order of a natural text description. Through our evaluations of the graphical model, we show that for different feature detectors, datasets, and graph sizes the graphical model helps improve the accuracy of the text output. We also show that the performance of the graphical model depends on the individual nodes (feature detectors) and that the model can be used to improve the performance of the feature detector. This thesis illustrates the whole process from images to text descriptors while evaluating components as we proceed. This work presents an approach to extract text labels from images using computer vision, a probabilistic graphical model, and natural language processing techniques
Memetic Algorithms, Domain Knowledge, and Financial Investing
No full textWhile the question of how to use human knowledge to guide evolutionary search is long-recognized, much remains to be done to answer this question adequately. This dissertation aims to further answer this question by exploring the role of domain knowledge in evolutionary computation as applied to real-world, complex problems, such as financial investing. The hypothesis is that domain knowledge can be combined with evolutionary algorithms (EAs) so as to systematically influence the traversal of a search space. A framework for incorporating domain knowledge into memetic algorithms, a specific kind of EAs, is proposed and is empirically evaluated by experiments that focus on knowledge-guided pre-processing, representation/initialization, and reproduction. Knowledge can be incorporated into an EA adaptedly (e.g., pre-processing and initialization) or adaptively (e.g., representation and reproduction). From the adapted perspective, knowledge can facilitate data collection and data cleansing in the preprocessing phase. Knowledge can also be utilized to create the initial population. From the adaptive perspective, individuals are represented by a sequence of components (or building blocks) that are mappable to the nodes in a semantic net, on which mutation and crossover are conducted. The mutation changes components to ones based on a certain semantic distance or more dynamically, based on a function of semantic distance. The crossover swaps the building blocks between two fit individuals, where the crossover point is constrained to be the boundary of these building blocks. The test problems are asset valuation and portfolio optimization. The experimental results show that knowledge can systematically influence the traversal of a search space. More interestingly, the results show how conceptual distance in human knowledge can correspond to distance between evolutionary individuals. According to the problems being solved, knowledge might be dynamically used to increase or decrease the reproduction size in a search algorithm. These results shed light on the role of knowledge in EAs and are part of the larger body of work in the field of evolutionary computation to delineate how domain knowledge might usefully constrain EAs
Synthesis of Functionalize Poly(propylene imine) Dendrons for a Multifunctional Gold Nanoparticle-Based Drug Delivery System for Advanced Pancreatic Cancer
No full textPancreatic cancer is the fourth leading cause of cancer-related deaths in the United States. According to the American Cancer Society, there is a five-year survival rate of only 5% once diagnosed with this deadly disease, and while surgery, radiotherapy and traditional chemotherapy are all options for the cure, they seldom do more than extend survival and relieve symptoms. Additionally, pancreatic cancer is often already at an advanced stage when diagnosed, mainly due to the asymptomatic character in the early stages of the disease as well as lack of early detection methods. Gemcitabine, the standard chemotherapeutic drug used to treat advanced stages of pancreatic cancer, only prolongs life marginally and relieves symptoms. It and other chemotherapeutic agents are characterized by their short half-lives and low therapeutic indices. While the administered chemotherapeutic drugs are meant for malignant cells, the drugs do not have the ability to specifically target them in vivo; therein lies the many unfortunate side effects due to the attack of normal cells. This research project is focused on targeted drug delivery. For this purpose, gold nanoparticles (GNPs) are used as the platform, since they enhance passive tumor targeting via the enhanced permeability and retention effect, and the transferrin (Tf) protein is used as an active cellular targeting moiety. As transferrin receptors are upregulated on the surface of 93% of pancreatic cancer cells, the transferrin protein is an excellent candidate as the targeting moiety. Herein, poly(propylene imine) (PPI) dendrons have been designed, prepared, and conjugated with Tf and used to decorate the surface of GNPs. Moreover, the conjugation of gemcitabine to PPI dendrons has been attempted through pH sensitive bonds, and other approaches are being considered for future work. The long-term goal of this research project is to produce a new type of multifunctional nanocarrier with the intention and expectation to increase the effectiveness of the delivered chemotherapeutic agents during treatment of patients with pancreatic cancer. This will presumably decrease the toxicity and drug resistance associated with conventional chemotherapeutic agents, as well as decrease the chemotherapeutic dose
A Study of Employee Assistance Programs in Higher Education Institutions
No full textIn this dissertation, a newer understanding of modern higher education institutional EAPs is discussed. The paper begins by explaining that the goals of EAPs are to keep employees with poor health from becoming ill and to improve health toward a positive state. EAPs are defined and broken down by type, including internal, external, and hybrid EAPs. Then, the difference in the mission and supervision between academic and corporate cultures is offered. The current focus of EAPs is on mental health services instead of the earlier focus on alcohol abuse treatment. In order to form hypotheses and help create the survey, pre-survey interviews were conducted with two EAP directors. The sample consisted of EAP directors that belonged to an organization (IAEAPE) dedicated to employee assistance excellence in higher education institutions. The overwhelming majority of the respondents reported having an internal EAP, whereas nearly half of the non-respondents had external EAPs. As hypothesized, modern EAPs in public and private institutions of all sizes are focusing on mental health services, with over 85% of the respondents reporting that they offered 6 or more mental health services. Also, 76% of EAPs reported having a staff member with a social work background. It was hypothesized that larger institutions would offer more services; instead EAPs from both small and large institutions reporting offering the same amount of services. EAPs in private institutions were more likely than public ones to offer more services, and both faculty and staff were more likely to use services more frequently in the private institutions. However, faculty use EAP services much less frequently than staff regardless of size or public/private status. Whites were far more likely than any other race to be reported by EAP directors as using their services frequently. Females were also more likely than males to use the EAP services frequently. Because these groups are not using services equally, it is important that EAPs market and tailor services for these infrequent users. The paper concludes with an assessment of the current state and future direction of EAPs
