Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
2 results
Search Results
Item FROM EXPLORATORY TO CONFIRMATORY: TOWARDS DATA VISUALIZATION AS A COMPLETE ANALYSIS TOOL(2023) Newburger, Eric C; Elmqvist, Niklas; Library & Information Services; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Confirmatory statistics tests, performed and written with equations, are a standard in scientific publications, but may represent a barrier to entry for novice analysts who have less familiarity with purely calculative methods. Data visualization, often touted as useful for sharing completed analyses with lay audiences, is often used for early-stage exploratory analysis. Could visualization support hypothesis confirmation? Do people have the visual intuitions to make use of such a tool? What would a visual statistical test look like, and what features would it require for acceptance by the scientific community?This research begins with a crowd-sourced experiment which asked respondents to fit a normal curve to a series of data samples, displayed as bar histograms, dot histograms, box plots, or strip plots. The results suggest people have visual intuitions – though biased toward overestimating spread – for linking idealized probability distributions with real sample data. People performed differently depending upon graphic form, suggesting design choices for subsequent experiments. A second experiment tested whether novice users might be able to perform a statistical test (T-Test) using a visual analogue – two overlapping distributions (shown as overlapping normal curves, box plots, strip plots, bar histograms, or dot histograms). Respondents had some capacity for this task, performing best with normal curves than with more detailed graphics like histograms. The final investigation of this research paired the design lessons garnered during experiments 1 & 2 with an interview study of experienced statisticians to explore the design requirements for creating acceptable visual tools for inferential statistics. The interviews uncovered three design foci: that the tool must display multiple, contrasting facets of analysis; the tool should connect the test back to raw data; and include a visual representation of real effect sizes compared to the p-value of the test statistic. The final chapter of this dissertation uses the design principles determined by these three investigations to propose a prototype visual tool for conducting a two-sample t-test, along with suggested variations for other inferential statistics.Item Multimodal Biomedical Data Visualization: Enhancing Network, Clinical, and Image Data Depiction(2017) Cheng, Hsueh-Chien; Varshney, Amitabh; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In this dissertation, we present visual analytics tools for several biomedical applications. Our research spans three types of biomedical data: reaction networks, longitudinal multidimensional clinical data, and biomedical images. For each data type, we present intuitive visual representations and efficient data exploration methods to facilitate visual knowledge discovery. Rule-based simulation has been used for studying complex protein interactions. In a rule-based model, the relationships of interacting proteins can be represented as a network. Nevertheless, understanding and validating the intended behaviors in large network models are ineffective and error prone. We have developed a tool that first shows a network overview with concise visual representations and then shows relevant rule-specific details on demand. This strategy significantly improves visualization comprehensibility and disentangles the complex protein-protein relationships by showing them selectively alongside the global context of the network. Next, we present a tool for analyzing longitudinal multidimensional clinical datasets, that we developed for understanding Parkinson's disease progression. Detecting patterns involving multiple time-varying variables is especially challenging for clinical data. Conventional computational techniques, such as cluster analysis and dimension reduction, do not always generate interpretable, actionable results. Using our tool, users can select and compare patient subgroups by filtering patients with multiple symptoms simultaneously and interactively. Unlike conventional visualizations that use local features, many targets in biomedical images are characterized by high-level features. We present our research characterizing such high-level features through multiscale texture segmentation and deep-learning strategies. First, we present an efficient hierarchical texture segmentation approach that scales up well to gigapixel images to colorize electron microscopy (EM) images. This enhances visual comprehensibility of gigapixel EM images across a wide range of scales. Second, we use convolutional neural networks (CNNs) to automatically derive high-level features that distinguish cell states in live-cell imagery and voxel types in 3D EM volumes. In addition, we present a CNN-based 3D segmentation method for biomedical volume datasets with limited training samples. We use factorized convolutions and feature-level augmentations to improve model generalization and avoid overfitting.