81 research outputs found
Doubly Robust Estimation of Causal Effects in Observational Data with Time-to-event Outcomes
No full textThere is a growing need for novel methods to estimate causal effects in observational data with time-to-event outcomes to provide reliable guidance for treatment strategies in life-threatening conditions. Bias in causal effect estimation can arise from treatment effect heterogeneity, model misspecification, and unmeasured confounders. This dissertation proposes novel approaches to address these challenges.
In the first part of the dissertation, I propose a framework for estimating conditional average treatment effects (CATE) in time-to-event data with competing risks. It accounts for treatment effect heterogeneity and protect against model misspecification. Using targeted maximum likelihood estimation (TMLE), I develop a substitution estimator based on cumulative incidence functions (CIF), derived from the efficient influence function (EIF). This estimator is doubly robust, and achieves asymptotic efficiency under mild conditions. Simulations demonstrate its favorable performance across various settings, confirm the double robustness, asymptotic normality and the flexibility of the framework incorporating different regression and machine learning models. Additionally, I construct variable importance measures to identify variables contributing to treatment effect heterogeneity and estimation, providing guidance for clinicians on the critical biomarkers or information to collect. The method is applied to electronic health record data to evaluate the treatment effect of steroids on ICU mortality among sepsis patients.
In the second part, I develop a novel instrumental variable (IV) method for estimating average treatment effects in data with unmeasured confounding. Derived from the EIF, this model-free estimator achieves double robustness and asymptotic efficiency under certain mild conditions. Defined by CIF, the method is adaptable to time-to-event data with competing risks. Our method also enables the incorporation of various models for outcome, treatment, and censoring. Extensive simulations demonstrate the double robustness, asymptotic normality, and the capability to analyze complex data. This proposed IV method is applied to investigate the effect of hydrocortisone on mortality among ICU patients with vasopressor-dependent septic shock.
Public health significance: The proposed methods address key challenges in estimating causal treatment effects in time-to-event data, including treatment effect heterogeneity, model misspecification, and unmeasured confounders. This dissertation provides powerful tools for optimizing treatment strategies, improving estimation reliability, and advancing healthcare research
Recommended from our members
Computational Approaches to Analyze and Interpret Rare and Low-frequency Genetic Variants
No full textThe growth of whole-exome sequencing (WES) and whole-genome sequencing (WGS) technologies has enabled efficient genome-wide profiling of rare and low-frequency genetic variants whose role in complex diseases remains unclear. Given the rich collections of data available in the past few years, we aim to understand the disease impact of these variants through computational approaches. First, we revisited a previous brain-related autism association signal from noncoding de novo variants in a WGS. We showed that this signal arises from a specific subset of variants and developed a new computational framework, Expression Neighborhood Sequence Association Study (ENSAS), to understand this signal by integrating gene expression correlations and sequence information, suggesting the signal’s potential linkage to synapse and chromatin states. Next, we analyzed the rare variants in a WGS cohort of bipolar disorder and schizophrenia cases and controls, performing functional annotation-based tests and genome-wide sliding window tests to identify groups of disease-associated rare variants. While restricted by current sample sizes and technical challenges we discovered a noncoding genomic region potentially associated with bipolar disorder, pending further validation. Finally, we developed MissenseHMM, a hidden Markov model-based approach that jointly models multiple missense pathogenicity scores and learns discrete states characterizing combinatorial patterns of these scores. We showed that MissenseHMM states demonstrate distinct enrichment for external genomic annotations and could enhance individual missense scores’ predictive power for distinguishing pathogenic and benign variants. Overall, our work provides new approaches useful for future analysis of genomic data
Identifying associations of de novo noncoding variants with autism through integration of gene expression, sequence, and sex information
No full textAbstract Background Whole-genome sequencing (WGS) data has facilitated genome-wide identification of rare noncoding variants. However, elucidating these variants’ associations with complex diseases remains challenging. A previous study utilized a deep-learning-based framework and reported a significant brain-related association signal of autism spectrum disorder (ASD) detected from de novo noncoding variants in the Simons Simplex Collection (SSC) WGS cohort. Results We revisit the reported significant brain-related ASD association signal attributed to deep-learning and show that local GC content can capture similar association signals. We further show that the association signal appears driven by variants from male proband-female sibling pairs that are upstream of assigned genes. We then develop Expression Neighborhood Sequence Association Study (ENSAS), which utilizes gene expression correlations and sequence information, to more systematically identify phenotype-associated variant sets. Applying ENSAS to the same set of de novo variants, we identify gene expression-based neighborhoods showing significant ASD association signal, enriched for synapse-related gene ontology terms. For these top neighborhoods, we also identify chromatin state annotations of variants that are predictive of the proband-sibling local GC content differences. Conclusions Overall, our work simplifies a previously reported ASD signal and provides new insights into associations of noncoding de novo mutations in ASD. We also present a new analytical framework for understanding disease impact of de novo mutations, applicable to other phenotypes
Light-Driven Flow: Laser Streaming and Optothermocapillarity
No full textThe manipulation of fluid motion through light-driven techniques has emerged as a promising area of research, offering innovative approaches to controlling fluid dynamics. This dissertation focuses on two aspects of light-driven fluid motion: laser induced acoustic streaming and laser-induced thermocapillary flow. Harnessing the energy of laser light, these two techniques present unique advantages, such as non contact manipulation, precise control over fluid behavior, and the ability to operate in micro-scale environments. Through comprehensive investigations, this work aims to unravel the complex mechanisms behind these phenomena, offering new possibilities for fluid manipulation in applications such as microfluidics, biomedical devices, and advanced manufacturing. First, the mechanism of laser-induced acoustic streaming (laser streaming) is explored. This light-driven method involves the use of a pulsed nanosecond laser on a metal launch pad, creating a jet-like motion from the surface. Experimental results reveal the presence of non-harmonic ultrasound pulse trains, which correlate with the velocity of the streaming jet. It is demonstrated that this jet motion is driven by acoustic streaming, propelled by the radiation pressure gradient from attenuated acoustic waves. The origins of these ultrasound waves through a combined analytical modeling and numerical simulation. The second part of this work addresses the challenges posed by laser-generated non-harmonic acoustic streaming. Traditional acoustic streaming empowered by harmonic pressure waves are incompatible with the miniaturized nature of modern microfluidics. Meanwhile, the prevailing analytical tool for conventional acoustic streaming, i.e., the progressive approximation (perturbation) method, becomes invalid for non-harmonic fields. A new numerical methodology is developed by decomposing the ultrasound signal into a Fourier’s series and making it possible to apply the progressive approximation method. Numerical modeling of the laser streaming field is presented. The final part of this work focuses on laser-induced thermocapillary flow, which can cause varied surface deformations (local depression/local elevation) in liquid layers. This study identifies geometric confinement as a critical factor in determining surface behavior. Furthermore, the competition between interfacial thermocapillary flow and returning flow within the liquid dictates whether the surface rises or falls. A criterion for distinguishing between "thin" and "thick" liquid layers is established, with experimental results aligning with theoretical predictions
A Study on the Change Path of Public Administration Mode in the Era of Big Data
No full textThe advent of the big data era presents social public administration with a host of novel opportunities and challenges. Harnessing the capabilities and guiding influence of big data in a scientific manner is essential for enhancing the current standards of social and public administration. This paper delves into the critical issues confronting the existing public administration model within the big data context, including information security concerns, challenges in information integration, and the lack of information awareness among staff. It proposes targeted reform measures aimed at fostering the evolution of the public administration model to better adapt to and thrive in the new era
Linking Instructional Leadership and School Support to Teacher Expertise: The Mediating Effect of Teachers’ Professional Development Agency
No full textThe focus on developing teacher expertise makes teaching and learning more sustainable, as it is a way of working to create improvement in education. The objective of this study was to explore the direct or indirect impacts of principal instructional leadership and school support on teacher expertise and explore the mediating effect of teachers’ professional development agency. A survey of 1123 teachers was conducted at 21 primary schools and 20 secondary schools in Hebei and Shanxi provinces of northern China. Structural equation modeling and bootstrapping were performed to test the relationships between variables. Results showed that teachers’ professional development agency mediated the effects of principal instructional leadership and school support on teacher expertise. School support was a better predictor of teacher expertise than principal instructional leadership. Providing instructional conditions and leadership support were non-significantly related to teacher expertise. Colleague support and student support were the better predictors of teacher expertise than providing instructional guidance and monitoring. The findings indicate that the growth of teacher expertise depends on building their professional development agency. Teachers will have a strong sense of agency to sustain the teaching profession when principals establish a supportive school climate that emphasizes teaching and learning in their leadership practice and enables teachers to build positive relationships with colleagues and students. The study confirms the supportive factors that impact teacher expertise and provides useful implications for the daily practice of teachers, principals, and administrators
Real-time Blind Deblurring Based on Lightweight Deep-Wiener-Network
No full textIn this paper, we address the problem of blind deblurring with high
efficiency. We propose a set of lightweight deep-wiener-network to finish the
task with real-time speed. The Network contains a deep neural network for
estimating parameters of wiener networks and a wiener network for deblurring.
Experimental evaluations show that our approaches have an edge on State of the
Art in terms of inference times and numbers of parameters. Two of our models
can reach a speed of 100 images per second, which is qualified for real-time
deblurring. Further research may focus on some real-world applications of
deblurring with our models.Comment: imcomplete figure
A Study on the Optimization Strategy of Teaching Models for Scientific Thesis Writing Courses for Undergraduates in Public Administration
No full textBased on the case of School S, this paper examines the current state of scientific thesis writing courses for undergraduates in the field of public administration. The research reveals several key issues, including insufficient course content, an irrational course structure, and a relatively monotonous approach to teaching methods. The authors propose several strategies to solve these problems. Such as boosting the richness of the course content, refining the course structure for a deeper understanding of the thesis writing process, and integrating a variety of teaching methods to stimulate students' interest as well as improve their writing skills
Clip as RNN: segment countless visual concepts without training endeavor
No full textExisting open-vocabulary image segmentation methods require a fine-tuning step on mask labels and/or image-text datasets. Mask labels are labor-intensive, which limits the number of categories in segmentation datasets. Consequently, the vocabulary capacity of pre-trained VLMs is severely reduced after fine-tuning. However, without finetuning, VLMs trained under weak image-text supervision tend to make suboptimal mask predictions. To alleviate these issues, we introduce a novel recurrent framework that progressively filters out irrelevant texts and enhances mask quality without training efforts. The recurrent unit is a two-stage segmenter built upon a frozen VLM. Thus, our model retains the VLM’s broad vocabulary space and equips it with segmentation ability. Experiments show that our method outperforms not only the training-free counterparts, but also those fine-tuned with millions of data samples, and sets the new state-of-the-art records for both zeroshot semantic and referring segmentation. Concretely, we improve the current record by 28.8, 16.0, and 6.9 mIoU on Pascal VOC, COCO Object, and Pascal Context
- …
