-
Prospective Students
-
Faculty and Staff
-
Current Students
-
Alumni and Friends
The present dissertation research aimed to identify factors that influence food purchase for families with young children, and further, to understand if family communication patterns play a role in the ways families communicate about food.
Three studies were conducted to address these aims. An online questionnaire focused on food purchase influences, food purchase discussion, and family communications patterns was administered to the primary food shopper for families with elementary school-aged children living in the United States. The first study utilized data from this online questionnaire to determine the factors most frequently influencing food purchase and to determine the association between income and the influence of certain factors on food purchase. The second study also used data from the questionnaire and identified the frequency of food purchase discussion within the sample, as well as the association between income and high levels of conversation. STATA 16 data analysis software was utilized to perform logistic regression and calculate unadjusted and adjusted odds ratios and 95% confidence intervals for each of the first two studies. The third study utilized qualitative interviews of questionnaire participants from each of the four family communication pattern types to collect data and thematic analysis was used to determine how families with children communicate about food or other purchases.
Participants in the online questionnaire reported availability, buyer preference, household preferences, and convenience as most frequently influencing their food purchases of both fruits/vegetables and “junk” food, or food generally considered non-health promoting. For fruits and vegetable purchases, there was a statistically significantly increased odds of marketing as an influence in households with income of less than $50,000 compared with households with an income of $100,000 or more (OR=9.01; 95% CI: 4.06, 20.01), and the association was attenuated, but remained statistically significant when adjusting for numerous demographic characteristics (AOR=7.56; 95% CI: 2.15, 26.52). Additionally, households in the lowest income group had statistically significantly increased odds of food quantity influencing fruit/vegetable purchases when compared to the highest income group, although findings were no longer statistically significant after adjustment for marital status, education, and age. (OR=2.66; 95% CI: 1.05, 6.74 and AOR=1.27; 95% CI: 0.26, 6.14). There were similar associations for “junk” food, in that the lowest income group had increased odds of marketing (OR=4.21; 95% CI: 1.91, 9.30 and AOR=2.21; 95% CI: 0.77, 6.36) and food quantity (OR=2.53; 95% CI: 1.16, 5.52 and AOR=4.05; 95% CI: 1.08, 15.23) influencing their purchasing when compared to the highest income group, although adjusted results were only statistically significant for the food quantity outcome. For “junk food”, the lowest income group also had increased odds of food beliefs influencing their purchase compared to those with incomes $100,000 or more (OR=3.02; 95% CI:1.39, 6.60 and AOR=2.42; 95% CI: 0.79, 7.43).
The second study, which utilized data collected from the online questionnaire, revealed that nearly 94% of participants reported discussing food purchase with their families, and over half (51%) of participants reported high levels of conversation orientation. Households with incomes $50,000-$99,999 had increased odds of reporting high levels of conversation compared to those with incomes of $100,000 or more (OR=2.43; 95% CI: 1.06, 5.57), and these odds remained similar in an adjusted model including both income and education (AOR=2.61; 95% CI: 1.23, 5.60).
The qualitative study reaffirmed that many families talk about food purchase and want children to understand the cost of items that are purchased. The family communication pattern types derived from the scale used in the questionnaire were often not evident based on responses given during the interviews. Reliance on the primary shopper to do all parts of the food acquisition process was a prominent theme derived from the interviews, and while some preferred this, others noted wishing they had more assistance from their partner or other family members. Participants also frequently noted the cost of food as a major concern more generally, and that price does play a major role in what they decide to purchase when shopping.
The findings from these three studies indicate that families are communicating about food purchase and identifies key purchasing influences. Regardless of a family’s communication about other topics, food purchase discussion may be an opportunity for children to influence the food that is brought home by the primary food shopper. Thus, health professionals should consider leveraging discussions of balanced diets with both children and caregivers to influence the purchase of healthy food options for consumption.
As the adoption of electric vehicles (EVs) grows, optimizing the EV charging process becomes crucial for efficient energy utilization and grid management. Some key aspects related to EV charging optimization include charging cost minimization, battery health and efficiency optimization, EV user behavior analysis, EV charging stations placement or location optimization etc. Through this dissertation, our aim is to contribute to the field of EVs in a way that helps in optimizing the huge infrastructure that will be required in coming years to accommodate the large number of EVs on roads.
In the first chapter, the problem under consideration is optimizing the charging schedule of an electric vehicle to minimize the total charging cost. In order to preserve battery life, the charging rate is restricted based on the current state-of-charge (SoC) of the battery. Noting that the charging rate limit is typically represented as a decreasing concave function of SoC, we first formulate the problem as a constrained optimal control problem. Discretization of the problem poses a nonlinear programming (NLP) problem that has a linear objective function and a convex feasible region. The proofs of the convexity of the feasible region and the strong duality of the problem are presented. Exploiting the strong duality, we present an exact solution approach that employs a cutting plane method to solve the Lagrangian dual problem in conjunction with the recovery of the primal solution. Subsequently, we propose two heuristics that employ greedy strategies, where charging is conducted to its rate limits over the periods with the lowest costs. We also present examples that illustrate these greedy strategies may not yield exact solutions. A thorough numerical experiment on simulated data is provided for the comparative efficacy of the proposed methods to the existing method.
In the second chapter, we consider a disjointly constrained bilinear program in which the variables are partitioned into two disjoint sets, each with its own respective set of constraints while bilinear terms in the objective function relate the two sets of variables. Leveraging the fact that the problem reduces to a linear program when one set of variables is held constant, we initially provide a geometric characterization of the relationship between the optimality condition of a basic feasible solution within one set of variables and the corresponding polyhedron in the other set of variables that achieves optimality, which we call the optimality polyhedron. By exploiting this relationship, we propose an algorithm that implicitly enumerates basic feasible solutions to expand the set of optimality polyhedra that eventually covers the feasible region in the other set of variables. A numerical study on randomly generated instances reveals that the proposed algorithm examines only 3.56% of the total number of basic feasible solutions on average while generating higher quality solutions compared to an off-the-shelf solver.
In the third chapter, we formulate the hard clustering problem of assigning a data point to exactly one cluster as a bilinear optimization problem. Using the Manhattan distance between points and their respective cluster centers as the objective function, the optimization problem is formulated as a minimization problem. This bilinear optimization problem is solved using the polyhedra expansion algorithm with initial clustering from k-means algorithm. The application of this methodology is demonstrated on the National Household Travel Survey data by utilizing the basic demographic and psychographic variables. The objective of this study is to develop a clustering scheme that can be utilized as a baseline to create potential customer segments by analyzing the driving behavior of a group of EVs.
β-ZnTe(en)0.5, an organic-inorganic hybrid material, demonstrates exceptional long-term stability exceeding 15 years in ambient conditions, surpassing materials like perovskites. This stability arises from strong covalent-like bonds within its quasi-2D layered structure, where ZnTe inorganic layers are bonded with ethylenediamine. The material exhibits quantum confinement effects, resulting in a significant bandgap blueshift compared to bulk ZnTe, and anisotropic thermal expansion with a low uniaxial thermal expansion coefficient. High crystallinity, evidenced by narrow Raman line widths, further highlights its quality, making β-ZnTe(en)0.5 a promising candidate for optoelectronic applications.
β-ZnTe(en)0.5 crystals were synthesized via a solvothermal method, leveraging high autogenous pressures and controlled crystallization in sealed reaction vessels at elevated temperatures. This technique, using ethylenediamine as a solvent, facilitated the formation of colorless flake-like crystals through controlled reaction parameters and purification. For device fabrication, a shadow mask approach was chosen over conventional lithography like UV photolithography and electron beam lithography (EBL), offering a resist-free, versatile, and less invasive patterning method that avoids chemical exposure and preserves material integrity. This allowed for high-quality devices with minimal artifacts, enabling reliable transport property investigation.
To explore charge transport, Space-Charge-Limited Current (SCLC) analysis, based on the Mott-Gurney law (J ∝ V²), was employed. While real materials deviate from ideal behavior, this analysis established a baseline understanding of charge carrier mobility. Two-probe electrical measurements on both vertical and lateral device configurations were conducted. Initial vertical measurements on pristine samples with metal electrodes revealed SCLC behavior, with mobilities of 8.8 × 10-3 cm2/(Vs) and 2.5 10-3 cm2/(Vs) for samples synthesized in 2007 and 2019, respectively. Lateral measurements, conducted along the a- and c-axes, revealed significantly higher mobilities, with values of 1.787 × 102 cm2/(Vs) along the a-axis and 1.696 × 101 cm2/(Vs) along the c-axis, demonstrating anisotropic charge transport. These results underscore the importance of SCLC measurements in characterizing the anisotropic transport properties of materials.
Complementary electronic structure analysis using X-ray photoelectron spectroscopy (XPS), ultraviolet photoelectron spectroscopy (UPS), Kelvin probe force microscopy (KPFM), and hot probe measurements were performed. XPS revealed a valence band maximum (Ev) of 0.80 eV below the Fermi level (EF), indicating p-type conductivity, corroborated by a 3.55 eV bandgap from photoluminescence (PL). KPFM yielded a work function of 4.59 ± 0.03 eV, consistent with UPS data. Integrating these results produced a reliable energy band diagram, confirming p-type conductivity and establishing band alignment. This multi-technique approach emphasizes its importance for accurate electronic structure determination.
In this dissertation, we discuss invisibility and inverse problems.
First, we discuss nonradiating orbital motions. We theoretically create nonradiating sources that orbit about the center of an annulus and are of whatever shape we want them to be. And we also discuss how one could possibly create an experimental setup demonstrating nonradiating orbital motions. The examples we discuss are all 2-D scalar wave problems.
Then, we discuss nonscattering scatterers. Building on the work of A. J. Devaney and others, we discuss about how to theoretically create objects that are invisible from some directions but not others. We find that, the more directions of invisibility our objects have, the harder they become to see when looking at them from between the object's directions of invisibility.
Finally, we discuss a globally convergent numerical method for solving a coefficient inverse problem. In particular, we discuss a Carleman-Picard iteration method for reconstructing the coefficient function for an inverse problem involving a parabolic PDE. The coefficient function can represent, among other things, a hidden object that we want to find without disturbing the medium in which our object is hidden. We demonstrate how well our new numerical method works with three tests.
We propose frameworks for dimension reduction in high-dimensional Vector Autoregressive (VAR) models using Spatial Quantile Regression (SQR). By incorporating adaptive Lasso and SCAD regularization, our methods enable robust inference under heavy-tailed or non-Gaussian errors while performing automatic
variable selection. To further address over-parameterization, we develop a tensor-based approach—Multilinear Low-Rank Spatial Quantile Regression (MLRSQR)—which restructures VAR transition matrices into low-rank tensors for simultaneous parameter reduction and quantile-wise modeling. Additionally, we
introduce the Sparse Higher-Order Reduced-Rank SQR (SHORRSQR) estimator, integrating Lasso penalties for sparsity, and design efficient ADMM-based algorithms.
Ultraviolet (UV) radiation in the 200–300 nm range is highly effective for microbial deactivation and has become a promising disinfection technology. There are several sources of UV radiation, including mercury vapor lamps, xenon arc lamps, deuterium lamps, and light emitting diodes (LEDs). Because of the associated damage to the skin, UV-LEDs are considered the safest and can provide a sustainable, energy-efficient, and environmentally friendly solution. However, there are some limitations in energy management, efficiency, automation, and optimization of UV-LEDs due to their wavelengths. Therefore, to optimize the efficiency in view of the application to microbial disinfection four UV-LED wavelengths (255 nm, 265 nm, 275 nm, 285 nm) were investigated. Thus, the dissertation work focused on the design, prototyping, and testing of an automated solar-powered UV-LED-based disinfection system, which serves as a sustainable and efficient solution for microbial deactivation in water, air, and surface applications.
Using the double-agar layer technique, the testing of these UV-LEDs for their ability to disinfect microbes/bacteriophages (MS2 and Phi6) was performed. Experimental results indicate that shorter wavelengths (255 nm) achieved the highest microbial inactivation with lower energy input, while longer wavelengths (285 nm) required significantly higher doses. The 255 nm UV-LED system, at a UV dose of 94.3 mJ/cm², achieved a 2.03 log reduction (99.08% reduction) for MS2, and at 117.9 mJ/cm², it achieved a 2.14 log reduction (99.28% reduction) for Phi6. In contrast, the 285 nm UV-LED system required 174.6 mJ/cm² for a 1.65 log reduction (97.78% for MS2) and 192.06 mJ/cm² for a 1.75 log reduction (98.25% for Phi6). The regression analysis and two-way ANOVA showed that the UV dose and wavelength are statistically significant in deactivating the microbial organisms. Finally, an automated solar-powered UV-LED-based disinfection system was designed, built, and tested. Similar results were obtained with the same statistical significance.
This research established solar-powered UV-LED disinfection systems as a scalable, sustainable, and highly effective solution for microbial decontamination. These findings will contribute to advancements in UV-LED technology, renewable energy integration, and automated disinfection, with diverse applications in healthcare, surface, water treatment, and public sanitation.
Exterior wood wall studs that serve as an integrated wall system in conventional house framing perform as very inefficient components when referring to the resistance of the conductive flow of heat (R-value). Affordable and available technology for wooden wall studs has not improved in decades, consistently resulting in lower R-values of the wall at the location of the studs as compared to the cavity between them. Typical wall stud sizes range from 2”x4” (1-1/2”x3-1/2”) to 2”x8” (1-1/2”x7-1/4) and net R-values of 4.38 and 9.07 respectively, while the insulation cavities would net R-13 and R-22 using standard fiberglass batts.
Construction techniques have evolved for outer surface areas of exterior walls by using various materials adhered to the outside of the wall surface. Modulating the actual stud on the interior, center, or outside with an insulative material could increase the R-value of the actual stud to a resistance of conductive flow of heat that is more in line with the insulated cavity. Aligning the R-values of the modified wooden stud and the insulated cavity could hypothetically increase the efficiency of all other exterior applied insulative materials, creating a higher thermal efficiency than is currently realized.
While teachers typically receive thorough training in their specific curricular areas through state licensure programs, they often do not receive adequate instruction on implementing evidence-based practices and interventions in general classroom settings (Mahoney, 2020). Consequently, there is an increasing concern that high school educators, in particular, may lack the knowledge and support necessary to effectively utilize evidence-based practices and interventions within a multi-tiered framework such as the Multi-Tiered System of Supports (MTSS). Educators who are not adequately trained or do not have a clear understanding of the implementation process and insufficient support from their school or district may struggle to maintain fidelity in MTSS implementation.
Implementing MTSS in high schools can be challenging due to many systemic factors. Leaders sometimes lack clear guidance due to the autonomy and flexibility permitted by different state education departments. This situation requires district and school administrators to navigate the complexities of selecting cost-effective resources, timely data-based decision-making, stakeholder collaboration, and the availability of qualified personnel to implement MTSS effectively (Clark & Dockweiler, 2019). This study utilized a multiple-case study methodology to explore the experiences of three Algebra I teachers and two English II teachers who implemented MTSS in their high schools across three districts in South Carolina. The findings underscore the importance of developing a strategic leadership plan that enables stakeholders to collaborate effectively and comprehensively understand the MTSS framework. The study findings emphasize a need to enhance teacher capacity throughout the MTSS implementation process by providing targeted professional development centered on innovative strategies and essential tools. Results offer valuable insights into the challenges high school English and Math teachers encountered during the implementation of MTSS at their schools and provide recommendations for school and district leaders aimed at creating clear and well-defined MTSS implementation plans.
Serendipity is a concept associated with accidental and unexpected discoveries that are valuable. In the recent decade, many researchers have advocated serendipity, as part of the “beyond accuracy” metrics, to encourage a recommender system to be an exploratory discovery tool instead of a narrowly focused machine. However, due to serendipity’s elusive and subjective nature, it is challenging to model. Collecting large-scale ground truth data is also a challenge. In this dissertation, I addressed both the challenges of serendipity model construction and the ground truth collection for recommender systems.
Leveraging the recent breakthrough in generative AI and large language models, I utilized three types of generative AI models: Large Language Models (LLMs), Transformers-based cross-domain models, and Diffusion Models (DMs), to construct a serendipity recommendation model. In addition, I used Large Language Models to collect serendipity ground truth data from large-scale e-commerce reviews data. The extensive experiments demonstrated the effectiveness of generative AI in modeling serendipity and ground truth collection. This dissertation advances the understanding and implementation of serendipity in recommendation algorithms, which will empower ordinary people with opportunities of bumping into unexpected but valuable discoveries.
Bulk RNA sequencing (RNA-seq) provides a cost-effective overview of gene expression but lacks resolution to identify cell-type-specific contributions in heterogeneous tissues. Computational deconvolution methods address this by estimating cell-type proportions from bulk data, enabling finer biological insights. This dissertation develops and applies statistical frameworks to improve the accuracy and interpretability of deconvolution results.
We begin by reviewing RNA-seq technologies and the impact of cellular heterogeneity. Deconvolution is then framed as a nonnegative matrix factorization (NMF) problem, with attention to challenges like non-uniqueness and noise sensitivity. Building on recent identifiability theory, we propose a geometric structure-guided NMF (GSNMF) that incorporates biological priors—such as marker genes—and local manifold structure to stabilize estimation.
To further enhance reference-free deconvolution, we introduce pseudo-bulk augmentation: a strategy that synthesizes single-cell-derived mixtures to enrich bulk data. This approach mitigates issues related to underdetermined solutions and improves robustness.
A comprehensive benchmarking study compares reference-based and reference-free methods using metrics like correlation, root mean squared error, and mean absolute deviation. Results show that while high-quality reference data can improve performance, augmented reference-free approaches like GSNMF are highly effective when reference data are scarce. We conclude with future directions and ongoing challenges.
