University of Maryland, Baltimore County
Not a member yet
17643 research outputs found
Sort by
Information gathered by retrospective, self-report, emotional frequency items in children
No full textRetrospective emotional frequency appraisals are often used in clinical assessment measures, but their suitability for use with children has not been well studied. The aims of this project were to (a) examine whether items that use retrospective frequency structures gather more or less information than items that do not use such structures and (b) examine whether the information gathered by such items differs across children's ages. Method. Data were gathered from 9- to 12-year-old girls who participated in a larger study of a depression treatment protocol. Two sets of five pairs of items were sampled from two children's depression measures. The item pairs contained one item from each measure. One set of item pairs was matched for content and the use of retrospective frequency structures. The other set was matched for content only. Results. For the first research question, information curves for the two item sets were generated using Samejima's (1969) Graded Response Model (GRM). Visual analyses of the information curves provided inconclusive results as to whether the presence of retrospective frequency structures is associated with differences in item information levels. The second research question was conducted in two parts. For both, only data from the 9- and 12-year-old participants were analyzed. In the first part, confirmatory factor analysis was used to analyze measurement invariance across the two groups' responses. Theses analyses showed signs of measurement non-invariance in both item sets. The second part of the analyses was conducted by generating separate GRM information curves for the two age groups and conducting visual analyses of the information curves. These analyses showed that the model which had been used throughout the remainder of the study did not fit the 9-year-old group well. They also showed that the 12-year-old group's information curves varied more in height across measures and item sets than did the 9-year-old group's curves. Discussion. Although the findings failed to shed light on the effects of retrospective frequency structures on children's responding, they highlighted potential differences between the 9- and 12-year-old groups' factor structures and indicated that the 9-year-olds displayed decreased sensitivity to differences in item structure
Feature Extraction and Fusion for Supervised and Semi-supervised Classification: Application to fMRI and LTM data
No full textExtracting powerful features from high dimensional noisy data promises to significantly improve the effectiveness of further analysis, especially of classification. Since there is no single feature selection and extraction method or classifier that works best on all given problems, developing effective and efficient feature selection and extraction methods and classifiers for specific applications has become one of the most active areas in the machine learning field. The aim of this dissertation is to develop novel data-driven methods for extracting and selecting the most distinguishing features for performing classification using functional magnetic resonance imaging (fMRI) and laser tread mapping (LTM) data. FMRI data have the potential to characterize and classify various brain disorders including schizophrenia. However, the high dimensionality and unknown nature of fMRI data present numerous challenges to accurate analysis and interpretation. Independent component analysis (ICA), as a data-driven method, has proven very useful for fMRI analysis in extracting spatial components as multivariate features used in classification, and more recently, for the analysis of fMRI data in its native complex-valued form. In this dissertation, we first present a novel framework to extract powerful features from components estimated by ICA, allowing us to remove the redundancy and retain the most discriminative activation patterns from multivariate ICA features. We apply the proposed three-phase feature extraction framework to two real-valued fMRI data sets, and achieve high classification rates in discriminating healthy controls from patients with schizophrenia. Second, due to the iterative nature of ICA algorithms, typically independent components (ICs) are not estimated consistently when running ICA multiple times, and hence it is not clear which result to use further. We present a statistical framework that utilizes an objective criterion to select the best of multiple ICA runs such that the multivariate ICA features from the best run can be used for further analysis and inference. Using the proposed framework, we study the performance of a novel complex ICA algorithm for fMRI analysis, entropy rate bound minimization (CERBM), which takes all three types of diversity into account, including non-Gaussianity, sample dependence and noncircularity that are present in the complex-valued fMRI data. We show that CERBM leads to significant improvement in ICs that provide high classification accuracy, and thus is a promising ICA algorithm for the analysis of complex-valued fMRI data. Classification using LTM data is another problem we address where we first study the use of highly multivariate solutions such as ICA and then note the advantages using lower-level features for classification. In this case, an important problem is the selection of best set of features for the best classification performance. Additionally, there is a large amount of unlabeled tire data that are easy to collect but only a few of them can be easily labeled by an expert. In this dissertation, we propose a novel mutual information (MI) based approach to achieve feature splits for co-training, a practical and powerful data-driven method in semi-supervised learning. Inspired by the idea of dependent component analysis, the proposed MI-based approach presents feature splits that are maximally independent between- or within- subsets, and thus selects and fuses features more effectively than other feature split methods. Experimental results from both simulations and LTM tire data indicate that co-training with the MI-based feature split yields significantly higher accuracy than supervised classification
The Impact of Computer Assisted Language Learning Adhering to the National Standards for Foreign Language Learning: A Focus on Modern Standard Arabic at the University Level
No full textThe development of the new digital world in the 21st Century has exponentially altered the mode through which we communicate, teach, learn, and perform research. The mode of interactions and social relationships that occur through new digital technologies are also shaping and transforming the meaning of instruction. Current online activities support face-to-face, distance education and hybrid or blended courses (combination of face-to-face and exclusively online interactions) (Garnham & Kaleta, 2002). Yet, despite the more frequent use of technology in the Arabic classroom, there is still little research conducted in the Arabic instructors' use, understanding, and knowledge of technological tools. As a result, this study examined the use of technological tools Arabic instructors are implementing in their curriculum. This study also explored Arabic instructors' awareness of the National Standards for Foreign Language Learning (NSFLL) and how they integrate technology to promote NSFLL in the teaching and learning of Arabic. The study also investigated the instructors' roles as online facilitators and their beliefs about the importance of the integration of technology to enhance Arabic learners' foreign language (FL) skills at university level in the United States. This study used a mixed-method approach to collect data from a sample of Higher Education Arabic instruction using two types of instruments. First was the online non-experimental survey, which was computer-mediated and self-administered survey. Second were in-depth semi-structured interviews conducted with a small sample of instructors to obtain an in-depth understanding of Arabic teachers' attitudes toward the use of technology and their beliefs concerning the NSFLL. The interviews investigated suggestions and recommendations of Arabic instruction to the National Standards for Foreign Language Learning in the virtual learning environment. The significance of the study was to provide contribution and support to the progress of the teaching and learning of Arabic as a FL in the United States in the 21st Century
Welfare Recidivism in Maryland: Does Child Support Make a Difference?
No full textThe purpose of this study is to identify associations between returns to welfare and the receipt of child support in Maryland. The sample used was limited to women who had a child under the age of 18 at the time of exit and current child support due during the three years after exit. The final sample was 42,234 welfare exits. Results suggest a significant association between receipt of child support and remaining off of welfare as well as the importance of the amount of child support received. Several limitations of this study and directions for future research are discussed
A Comparison of Sexual Attraction, Orientation, and Activity in Predicting Drinking Behaviors among Lesbian and Bisexual Females in the United States
No full textyour words Lesbian and bisexual women's attraction, orientation, and activity were examined in an effort to understand how sexualities are developed. Also, attraction, orientation, and activity were measured in relation to the binge drinking behaviors. Hypothesis (1), the three variables attraction, orientation, and activity are similar, but reveal distinctions. Hypothesis (2), compared to heterosexual women, lesbian and bisexual women will have more unhealthy drinking behaviors, especially bisexual women due to sexual minority stress. The National Survey of Family Growth (NSFG 2006-2010) public female and ACASI female data files were used. Results indicated that attraction, orientation, and activity are all separate factors that contribute to the development of a woman's sexuality. It was found that lesbian and bisexual women, especially bisexual women are more likely to engage in binge drinking behaviors compared to heterosexual women. In addition, attraction and activity appear to be slightly greater predictors of binge drinking behaviors
Computational Methods in Finite Mixtures using Approximate Information and Regression Linked to the Mixture Mean
No full textFinite mixture distributions are used in applications because of their ability to support heterogeneity. They also present interesting analytical challenges, often requiring special consideration in the selection of an appropriate model, inference of unknown parameters, and identifiability. The main contributions of this thesis are providing an approximation to the information matrix of a finite mixture of an arbitrary member of the exponential family, and a novel extension of the generalized linear model (GLM) with an underlying finite mixture distribution. Our approximation is equivalent to a complete data information matrix, which helps to explain previously noted connections between approximate scoring and the Expectation Maximization (EM) algorithm, and is further generalized to mixtures of an arbitrary member of the exponential family. To obtain convergence between exact and approximate information requires a clustered sampling assumption so that observations are sampled from the same (unknown) subpopulation of the mixture, providing an analogue to trials of a multinomial observation. We also consider a logistic regression model using a binomial finite mixture, so that the regression model is linked to the mixture mean. This significant extension of GLM appears promising for many potential applications such as modeling overdispersion in the data. Because the mixture mean is a composite parameter which does not appear explicitly in the likelihood, model formulation and inference pose both theoretical and computational challenges. We propose a random effects model with effects drawn from a set representing enforcement of the link. Initial results show that the model is effective at capturing extra variability while supporting the regression of interest
A Rapidly Deployable Image Classification System Using Feature Views
No full textConstructing an image classification system using strong, local invariant descriptors is both time consuming and tedious, requiring much experimentation and parameter tunings to obtain an adequate performing model. Furthermore, training a system in a given domain and then migrating the model to a separate domain will likely yield poor performance. As the recent Boston Marathon attacks demonstrated, large, unstructured image databases from traffic cameras, security systems, law enforcement officials, and citizens can be quickly amassed for authorities to review; however, reviewing each and every image is an expensive undertaking, in terms of both time and human effort. Inherently, reviewing crime scene images is a classification task. For example, authorities may want to know if a given image contains a suspect, a suspicious package, or if there are injured people in the photo. Given an emergency situation, these classifications will be needed as quickly and accurately as possible. In this work we present a rapidly deployable image classification system using “feature views”, where each view consists of a set of weak, global features. These, weak global descriptors are computationally simple to extract, intuitive to understand, and require substantially less parameter tuning than their local invariant counterparts. We demonstrate that by combining weak features with ensemble methods we are able to outperform current state-of-the-art methods or achieve comparable accuracy with much less effort and domain knowledge. We then provide both theoretical and empirical justifications for our ensemble framework that can be used to construct rapidly deployable image classification systems called “Ecosembles”. Finally, we recognize the fact that image datasets give us the relatively unique opportunity to extract multiple feature representations through the use of various descriptors. In situations where the original dataset is not available for further feature extraction or in cases where multiple feature views are ambiguous (such as predicting income based on geographical location and census data) the Ecosemble method cannot be applied. In order to extend Ecosembles to arbitrary datasets of diverse modalities, we introduce artificial feature views using kernel approximations. These artificial feature views are constructed from a single representation of the data, alleviating the need to explicitly extract multiple feature views. We then apply artificial feature views to a diverse range of non-image classification datasets to demonstrate our method is applicable to multiple modalities, while still outperforming current state-of-the-art methods
Constructing Change that Lasts: A Grounded Theory Study of Community-Based Arts' Creation of Social Impacts
No full textThe body of literature concerning social impacts of the arts, including research substantiating individual-level outcomes of arts participation, has grown a great deal in recent years, as has the Community Arts field's pursuit of more rigorous and useful evaluation approaches in light of challenging, contemporary demands on organizations. However, extant research has not fully answered how community-based arts organizations (CBAOs) conceptualize their pursuit of outcomes, what mechanisms underlie those pursuits, and how this translates into external impacts. In order to help fill gaps in the literature and contribute to evaluation efforts, this study applied a community psychology approach and utilized Constructivist Grounded Theory (Charmaz, 1995) to build upon a previous study that explored one CBAO's conceptualization and enactment of its goals (Scheibler, 2011). Through a close, multi-phased analysis of pre-collected (n=7) and newly collected interviews (n=11) of long-term and former participants of three representative CBAOs, the present study pursued new understandings of how participants' subjective experiences of program-fostered change processes convert to external and potentially longer-lasting impacts. This research revealed how participants were engaged by strengths-based program structures that fostered sense of community; how they were impacted by four change mechanisms: 1) fostering healthy maturation, 2) developing professional competencies, 3) building a creative foundation, and, 4) promoting change agent characteristics; and how transformative meaning-making enabled them to form new understandings of themselves, others, and society, which may enable them to be critical, productive, and life-long learners who can enact change in their communities
Remote sensing of vegetation structure using computer vision
No full textHigh-spatial resolution measurements of vegetation structure are needed for improving understanding of ecosystem carbon, water and nutrient dynamics, the response of ecosystems to a changing climate, and for biodiversity mapping and conservation, among many research areas. Our ability to make such measurements has been greatly enhanced by continuing developments in remote sensing technology - allowing researchers the ability to measure numerous forest traits at varying spatial and temporal scales and over large spatial extents with minimal to no field work, which is costly for large spatial areas or logistically difficult in some locations. Despite these advances, there remain several research challenges related to the methods by which three-dimensional (3D) and spectral datasets are joined (remote sensing fusion) and the availability and portability of systems for frequent data collections at small scale sampling locations. Recent advances in the areas of computer vision structure from motion (SFM) and consumer unmanned aerial systems (UAS) offer the potential to address these challenges by enabling repeatable measurements of vegetation structural and spectral traits at the scale of individual trees. However, the potential advances offered by computer vision remote sensing also present unique challenges and questions that need to be addressed before this approach can be used to improve understanding of forest ecosystems. For computer vision remote sensing to be a valuable tool for studying forests, bounding information about the characteristics of the data produced by the system will help researchers understand and interpret results in the context of the forest being studied and of other remote sensing techniques. This research advances understanding of how forest canopy and tree 3D structure and color are accurately measured by a relatively low-cost and portable computer vision personal remote sensing system: 'Ecosynth'. Recommendations are made for optimal conditions under which forest structure measurements should be obtained with UAS-SFM remote sensing. Ultimately remote sensing of vegetation by computer vision offers the potential to provide an 'ecologist's eye view', capturing not only canopy 3D and spectral properties, but also seeing the trees in the forest and the leaves on the trees
DEVELOPMENT OF ULTRA-MICRO ELECTRODE BASED BIOSENSORS
No full textHere I present a novel method for fabrication. characterization and optimization of micro scale, electrochemical aptamer-based sensors with the aim of developing sensors capable of quantitatively monitoring small molecule release from single cells in complex media. As a proof of concept, I have selected adenosine triphosphate (ATP) as a target molecule, since it is implicated in astrocyte communication in the central nervous system. Electrochemical methods provide powerful tools for the detection of molecular messenger release, but they are limited to molecules that are electrochemically active in a reasonable potential window. As such, current electrochemical methods for neuronal studies cannot detect ATP, since it is not electrochemically active in the preferred window for analysis. While the goal is to develop ATP sensors, the sensors developed can be generalized to any target molecule. The development of these sensors will enable study of gliotransmission with unprecedented spatiotemporal resolution and chemical specificity for analysis