22484 research outputs found
Sort by
Interpretable deep learning for advancing precision oncology
No full textThe goal of precision oncology is to provide each patient with the most appropriate cancer treatment. This approach involves understanding the impact of genomics on individual cancer cases and utilizing this understanding to tailor cancer therapies targeting the genetic mutations that drive the malignancy. Achieving this level of personalization in cancer treatment can be realized by integrating artificial intelligence (AI) to predict how a patient's cancer cells will react to anticancer medications based on their genomic data.
Three deep learning methods were developed for drug sensitivity prediction: 1) We used graph-regularized matrix factorization to decompose the drug response matrix into vectors for cell lines and drugs and developed mapping functions to convert observed molecular characteristics into these vectors. 2) A model was created to learn a mapping function from Somatic Genetic Alterations (SGAs) to gene expression. The hidden representations between these observed datasets served as the features of cell line to predict drug response. 3) An extended version of the Variational Autoencoder (VAE) model was utilized to condense gene expression and SGAs data into a low-dimensional, informative representation of cell lines. These representations were subsequently used for predicting drug sensitivity, replacing the raw omics data.
While all three models leverage representation learning for cell lines and subsequently apply the representation for drug sensitivity prediction, notable differences distinguish them. The first model excels in prediction performance, accurately forecasting drug responses for unseen drugs and cell lines. However, it falls short in terms of interpretability, which refers to how easily humans can understand its decisions. It involves the transparency of the model's internal workings, making clear how it processes input data to produce output. Conversely, the second model, though interpretable, falls short in performance due to its reliance on SGAs for predicting drug responses. To achieve a balance between predictive efficacy and interpretability, we developed the third model which takes gene expression and SGAs as input, utilizing a Deep Generative Model (DGM) and self-attention mechanism to enhance interpretability. All these three approaches contribute to the broader evolution of cancer medicine by aligning predictive accuracy with interpretive insights
Increasing Middle School Student Participation in IEP Meetings
No full textFederal and state guidelines on when students with disabilities should become more active in their own special education are notoriously vague, often leaving many school districts across the United States to their own interpretations on the appropriateness of when and how to do so. This is especially pertinent for younger students with disabilities, as federal law offers more guidance for those transitioning into high school. As a middle school special education teacher, I have repeatedly found that many of my students with disabilities who qualify for special education do not actively participate in their own Individualized Education Plan (IEP) meetings and are often wholly unaware that such a plan even exists. These observations, in the foreground of a host of relevant literature on the topic, have subsequently led to my growing concern that students who are not attending their own IEP meetings are in turn, limiting their opportunity to have an important stake in their own education through self-advocacy and self-determination, both of which are desired organizational goals at my school.
The purpose of this dissertation in practice is to explore the effectiveness of participation in IEP meetings amongst students with disabilities at the middle school level. Utilizing the methodological framework of improvement science, this study included the direct instruction of a modified curriculum designed to increase IEP meeting engagement and data collection from multiple observation points across two mock and one actual IEP meeting sessions. This theory of improvement is grounded in the notion that if given the information needed to attend and participate, middle school students with disabilities will become active participants in their own IEP meetings and take increased ownership over their own special education
Promoting Equity-Focused Math in Dual Language
No full textPublic Schools across the US face a common challenge in supporting minoritized students reach high academic standards. Standardized assessments repeatedly demonstrate Latinx and English Learner populations are underserved by US schools, reflecting the need to improve equitable access to high-quality instruction with a focus on diverse and multilingual learners.
Dual Language (DL) models have potential to address opportunity gaps, building upon students’ assets by developing academic bilingualism and biliteracy. This dissertation in practice focuses on improving Latinx students’ and English Learners’ access to effective mathematics instruction in a DL model. Using an Improvement Science framework, I facilitated a professional development sequence for DL math teachers in three cycles of Plan, Do, Study, Act (PDSA).
In each cycle, teachers were presented with specific techniques recommended by the National Council of Teachers of Mathematics (NCTM) and aligned to Guiding Principles for Dual Language Education. To determine how the intervention impacted instruction, I analyzed qualitative data collected through classroom observations and teacher interviews.
Teachers were more likely to implement concrete, visible strategies connecting mathematical representations. Students infrequently engaged in math dialogue, explaining reasoning, or persevering through struggle. As for DL strategies, most teachers incorporated language objectives and a variety of student groupings. However, few lessons included purposeful cross-linguistic connections, though these were more common in Spanish math instruction. New teachers were least likely to implement newly-learned practices.
Teacher interviews revealed specific challenges impacting their use of new strategies. New teachers were overwhelmed, and expressed needing more planning time. Lack of comfort with math content and curriculum were also barriers. Interviews and observations confirmed that classroom management influenced teachers’ infusion of effective math and DL practices.
Findings from this dissertation have important implications for leaders and educators. Aligning with prior literature, this DiP reinforces that student behaviors indicative of deep math learning require shifts from teacher-directed to student-centered instruction. Further, it is important to foster math classroom cultures that encourage students to make mistakes, celebrate perseverance, explain their thinking, and view struggle as part of learning. Establishing these norms for all students can interrupt the traditional teaching practices perpetuating disparities in math outcomes
Pleasure as Evaluative Perception
No full textPleasure is a familiar and normal part of everyday life. Consider the pleasure of stepping into a warm bath. There are at least three seemingly intertwined features of this experience. The first is perceptual: the experience involves the sensory feeling of the warm water on the skin of one’s legs. The second is evaluative: the warm water feels good on the skin. The third is motivational: one’s desire to submerge oneself fully in the bath. Ancient philosophy typically incorporated all of these elements in their discussions of pleasure, thereby furnishing unified and systematic accounts. Rooted in and inspired by Aristotle’s account of pleasure, I advance a novel framework with which to investigate foundational questions about pleasure. According to what I call the evaluative cognition framework, pleasure is best understood not as an object or property of experience, but as a form of evaluative perception—a way of sensing that something is good for the perceiver. Experiencing something as pleasant is a way of finding it good without necessarily thinking that it is good. A striking consequence of this framework is its rejection of a commonly accepted axiom in value theory: the claim that pleasure is intrinsically good. That the evaluative cognition framework leads to the rejection of this axiom may at first glance seem to warrant dismissing the framework out of hand. However, I contend that understood as a mode of evaluative cognition, pleasure can still play a foundational role in how we understand the relationship between value and action. Pleasure is a primitive form of valuing for human and non-human animals. It is best understood as a biological mechanism which, when working properly, leads animals to things that are genuinely good for them
Essays on Housing and Consumer Finance
No full textThis dissertation consists of three essays that contribute to the fields of housing economics, financial economics, and consumer finance. Chapter 1 charts the evolution of housing prices in American cities across the twentieth century. We use hedonic methods to construct annual market price indices for both rented and owned housing at both the national and city levels from 1890 to 2006. Using these indices, we document several new facts about housing markets, including, the relationship between housing market cycles and business cycles, the city- and national-level return to owning housing, the implications of a revised CPI on the standard of living, and the relationship between changes in housing prices and supply constraints. Chapter 2 combines hedonic pricing, machine learning, and text analysis methods to improve the predictive power of hedonic pricing models and reduce bias in hedonic indices. I use machine learning methods to select features from a novel geocoded, tabular dataset which I include as controls in hedonic pricing models to construct market price indices for both rented and owned housing at the city and sub-city levels for San Francisco from 1890 to 2006. Using these data and indices, I document the total return to owning housing at both the city and sub-city levels and dynamically compare the variance of sales price residuals within neighborhoods of San Francisco and across cities of the San Francisco Bay Area. Chapter 3 studies how the terms of consumer debt contracts affect nonpayment decisions of financially distressed borrowers. Using a proprietary credit report panel, we document that a substantial subset of financially distressed borrowers who hold a varied debt portfolio avoid defaulting on revolving credit, at the cost of an increased likelihood of future student loan default. We complement this finding with a survey that investigates the mechanism underlying these nonpayment decisions. We find that participants rank the timing of default as the most influential factor and link it to a low level of financial literacy for student loan default consequences
Wavelength Dependent Coherent Phonon Excitation near the E0 + Δ0 Critical Point in GaAs[001]
No full textI present a study of the wavelength dependence coherent phonon response of n and p-doped GaAs around its E0 + Δ0 critical point. When the excitation energy is above the fundamental band gap of GaAs (1.42 eV), excited charge carriers generate e-h plasma in the lattice, that can couple with the coherent longitudinal-optical (LO) phonons resulting in plasmon-phonon coupled modes. This response of the lattice and coupling with the charged carriers is observed using a pump-probe transient electro-optical reflectivity spectroscopic technique. I use a non-collinear optical parametric amplifier (NOPA) as a highly tunable excitation laser source with ~ 20 fs pulses for the transient reflectivity experiments. The femtosecond resolution makes it possible to follow the time evolution of the LO phonon and the plasmon-phonon coupled modes. I focus on the carrier-lattice response when excited across the E0 + Δ0 critical point (1.76 eV) in its band structure. Experiments are done using a wavelength range of 680 - 740nm (1.68eV – 1.83 eV). This wavelength range not only excites across the spilt-off band of GaAs but also allows the possibility of intervalley scattering from Γ valley into the L valley. As the scattering into the L valley becomes possible, a decrease in the phonon response is observed. An increase in phonon amplitude is seen when excitation energy is above split-off band gap. As excitation energy is further increased, a decrease in phonon amplitude can be seen again as scattering into L valley becomes more favorable.
To better understand the experimental trends simulations were done using the basic plasmon-phonon coupling equations. The frequency and amplitude of the plasmon-phonon coupled modes depends on the carrier mass and density. With the increase in carrier density the damping rate of the coupled modes initially goes up before eventually decreasing. This point of inflection can shift with change in carrier mass or plasmon frequency, thus being different for carriers in different valleys or bands. From the experimental data and using the insight gained from the simulations I present a qualitative picture for the wavelength dependent results
The impact of primary care nurse practitioner autonomy on workforce outcomes across various team compositions
No full textBackground: Nurse practitioners (NPs) are expected to take a central role in the future delivery of primary care working as autonomous providers within multidisciplinary teams. However, given the challenging work environments that often exist in primary care, it is imperative to understand the impact of their autonomy and teamwork on workforce outcomes such as burnout, job satisfaction, and turnover intention. While greater NP autonomy may lead to better workforce outcomes, effective teamwork can further improve these outcomes. However, few studies have examined the impact of NP autonomy and teamwork on NP workforce outcomes.
Purpose: We sought to examine the extent to which NP autonomy, NP teamwork, and their interaction are associated with NP workforce outcomes in primary care practices.
Methods: For our analyses, we used survey data obtained from 1,244 primary care NPs working for 1,109 practices in six states (Arizona, New Jersey, Washington, Pennsylvania, California, and Texas). Firstly, we conducted structural equation modeling (SEM) to estimate the associations between NP autonomy, measured by NP panel management, and NP workforce outcomes, with workload, measured by work hours, serving as a mediating variable. Secondly, we performed latent class analysis (LCA) to identify different primary care team compositions (an important determinant of teamwork) in which NPs work, using a survey item where NPs selected healthcare providers and staff they considered team members. We then assigned NPs to each distinct team composition based on their modal posterior probability and fit probit regression models to compare workforce outcomes among NPs in various team compositions. Lastly, we fit probit regression models to determine whether team composition identified in the previous step moderated the relationship between NP autonomy and workforce outcomes.
Results: 1) In SEM, fully autonomous management led to more burnout than co-managing (B=0.089, bias-corrected 95% bootstrap confidence interval [0.028, 0.151]). Work hours partially (27%) mediated this relationship. There were no significant differences in job satisfaction or turnover intention depending on the degree of autonomy in NP panel management. 2) Based on the LCA model fit evaluation, we selected a five-class solution. Conditional probabilities indicated that NPs in teams with frequent and extensive contact with other providers and staff, along with sufficient ancillary support, reported better workforce outcomes compared to those in teams with minimal contact with other providers and staff and limited ancillary support. However, results from the fully adjusted probit regressions suggested that workforce outcomes did not significantly differ across different team compositions. Further investigation revealed that the NP work environment, as an individual covariate, most effectively mitigated the influence of team composition on NP workforce outcomes. 3) The results of moderation analyses indicated that NPs with greater autonomy in panel management were more likely to experience poorer workforce outcomes, such as increased burnout and turnover intention, when working in teams compared to practicing alone.
Implications: NPs with greater autonomy in panel management tend to experience poorer workforce outcomes, and these outcomes worsen when they work in teams compared to practicing alone. Favorable work environments for NPs in specific team compositions can improve NP workforce outcomes. Specifically, for NPs with greater autonomy in panel management, optimal teamwork and additional support beyond the team may improve NP workforce outcomes and maximize the care provided by these NPs. Future research should incorporate patient characteristics and explore team processes across various primary care team compositions to gain deeper insights into the workforce outcomes for NPs with greater autonomy in primary care
From Hands-On to Hands-Free: It’s Time to Automate Your Data Displays
No full textPurpose & Goals
This poster introduces approaches to automating regular dashboard updates, driven by the need to improve upon workflow processes that have traditionally been performed manually. Our objective was to refine the mechanism which updates the Assessment unit's essential dashboards that showcase our collections and services and to develop a workflow that efficiently manages the influx of ad hoc report requests. Through automation, we sought to shift librarians’ focus from labor-intensive tasks to more meaningful data analysis, thereby optimizing efficiency and increasing the depth of insights gained.
Design & Methodology
An investigation into the automation of report updating focused on two primary tools: Microsoft PowerAutomate and Alteryx. To explore how the individualized features of each of these tools may affect their capability to cater to differing needs, dataset sizes, and resources, this project was conducted at two different institutions: Duquesne University and the University of Pittsburgh. At Duquesne University, PowerAutomate was deployed to automate the updates of hourly head counts. This was crucial for a mid-sized institution like Duquesne, which needed to monitor library space usage for staffing and operational hours. Data was originally collected via a LibInsight web form, with student workers entering time and user numbers by floor. The library analytics team would then download and publish this data monthly via PowerBI. To address specific needs, such as monitoring usage trends during finals, the frequency of data updates was increased to once daily. The adoption of a new automation approach transformed this process by enabling the direct parsing of data from LibInsight into SharePoint spreadsheets, automating data cleaning, and subsequently refreshing PowerBI. This fully automated workflow significantly reduced the requirement for active effort and streamlined the update process, operating seamlessly without manual intervention. On the other hand, the University of Pittsburgh, which encompasses several regional libraries and centers, had interest in exploring a greater diversity of tools due to challenges faced due to existing user familiarity with the available software. This led to the exploration of new tools, including Alteryx, which was timely introduced to the institution and found to be promising for updating dashboard data. Our initial trials included parsing data from Alma Analytics via API, data cleaning, and automatically updating Tableau datasets. Despite the considerable effort required for the initial setup, this approach has greatly streamlined data processing and analysis.
Findings
The exploration of automation tools has not only illuminated their potential to enhance operational efficiency but also emphasized their profound impact on our comprehension and handling of data. The development of these automated workflows has led to a thoughtful reassessment of our daily data management practices, encompassing activities such as recording, collecting, parsing, manipulating, and disseminating data. The automation of these routine tasks positions librarians to potentially devote more time to data analysis, thereby fostering more informed decision-making and facilitating strategic advancements.
Action & Impact
This project is an ongoing endeavor to refine and enhance the automation workflows within library systems. While we have garnered some insightful findings and reflections, the nature of this project is evolutionary, with continuous improvements and adaptations based on emerging data and feedback.
Practical Implications & Value
Libraries across the spectrum employ comparable systems, tools, and methodologies for data sharing and management, making the dissemination of practical use cases of automation workflows highly beneficial. Sharing these can significantly aid those entrenched in data-related tasks who aim to reduce time spent on routine operations and enhance their analytical endeavors. The specific workflows utilized by the authors for this project will also be made available, fostering a culture of efficiency and innovation within the library community. By adopting and adapting these proven workflows, the library assessment community is better positioned to concentrate on strategic analysis and decision-making
Acoustic Streaming by a Pinned Oscillating Membrane, Antifouling PDMS-SB, and Application to Microfluidic Artificial Lung
No full textAcoustic microstreaming in microfluidic devices is often used as a means to disrupt the typically laminar flows seen at the micro-scale for various purposes such as mixing and propulsion. Lower, audible frequency configurations can be cheaper and easier to generate, but are more limited in transduction methods due to long acoustic wavelengths. Existing audible frequency methods such as bubble and sharp edge streaming suffer from disadvantages of stability and obstruction.
In this dissertation, a novel configuration to generate acoustic streaming at audible frequency in a microchannel by a pinned oscillating membrane is detailed. Advanced characterization methods and computational fluid dynamics simulation show similar pattern and magnitude, providing evidence that this membrane oscillation is the driving mechanism of the time-averaged vortices. This method has potential application to the microfluidic artificial lung due to the vertical orientation of the resulting mixing allowing for augmentation of gas exchange across the permeable membrane, lack of any obstruction, and stability. Successful augmentation of gas exchange is initially demonstrated as shown by characterization of CO2 transferred into the channel. Scaling up of throughput is also demonstrated with a branching design, featuring a multilayer manifold to avoid undesirable interaction of the streaming flow with the channel geometry.
Also towards the microfluidic artificial lung, development of a modification of PDMS, called PDMS-SB, is described, which offers improved hemocompatibility, reduced coagulation, and reduced biofouling by surface deposition. The PDMS-SB is able to be easily integrated into fabrication processes designed for commercial PDMS as shown by compatibility with standard techniques such as spin coating, soft lithography, and plasma bonding, as well as showing similar mechanical properties and maintaining permeability.
Finally, blood testing experience with the PDMS-SB project was leveraged to characterize reduced platelet deposition and coagulation in microchannels with the actuated membrane due to physical motion of the membrane and fluid and taller allowable channel height reducing shear and preventing blockage. Successful augmentation of O2 transfer into ovine blood with increasing actuation strength is shown, reaching a biological target of oxygenation despite much lower surface area to volume ratio compared to other early microfluidic artificial lung works
Psychological Threat and Problem Solving in Physics: Relations and Effects of Mindfulness Training as an Intervention
No full textThe first aim of this work was to better understand the relationship between psychological threat—defined as perceived situational demands exceeding perceived coping resources—and accuracy, learning, and perceptions during problem solving. The second aim was to test the effect of mindfulness training on this relationship in the context of undergraduate introductory physics, an environment in which challenges relating to student motivation, retention, and equity are well known. One hundred and forty-nine undergraduates reporting psychological threat in physics were randomly assigned to receive either a 5-day mindfulness training intervention or no training (control). Both groups completed physics problem solving tasks before and directly after the intervention. Accuracy on three types of problem solving as well as momentary perceptions of confidence, anxiety, and difficulty were measured during the tasks at both timepoints. Learning on a preparation for future learning task was also measured after the intervention. Prior to the intervention, psychological threat was positively associated with perceptions of difficulty and anxiety, and negatively associated with perceptions of confidence and accuracy on quantitative and qualitative problem solving. Students assigned to mindfulness training reported lower psychological threat during the intervention, and greater confidence and lower perceptions of difficulty at posttest. However, reduction in psychological threat (measured during the intervention) did not mediate the effects of mindfulness training on confidence and difficulty at posttest. Mindfulness training led to a reduction of anxiety during problem solving for female and non-binary identifying students, and this effect was mediated by psychological threat. We found no effects of mindfulness training on problem solving accuracy or learning outcomes at posttest. We discuss the implications of these results for theories of psychological threat and problem solving