Four University of Washington Biostatistics Department doctoral students and one recent graduate were among the 2020 American Statistical Association (ASA) Student Paper Competition best paper award winners.
The doctoral students each received a $1,250 cash prize and all were invited to present at the 2020 Joint Statistical Meeting (JSM) in Philadelphia in August. JSM draws more than 6,500 participants and is the largest gathering of statisticians and data scientists held in North America.
Serge Aleshin-Grendel, a third-year doctoral student, was honored by the Social Statistics, Government Statistics, and Survey Research Methods section of ASA for his paper “Multifile Record Linkage and Duplicate Detection Via a Structured Prior for Partitions.”
The paper proposes a novel Bayesian approach to jointly perform record linkage and duplicate detection for an arbitrary number of files, so that rather than developing a new method each time one encounters a new setting, the paper’s method can be applied to new settings in a straightforward manner. A key feature, absent in previous approaches, is a mechanism for incorporating relevant prior information one may have about the data collection processes of the files, such as how much overlap one believes there is between files, or how much duplication one believes there is in each file.
Doctoral student Lucy Gao was honored by the Statistical Learning and Data Science section of ASA for her paper, “Testing for Association in Multi-View Network Data.”
The paper developed a statistical hypothesis test that can be used to determine whether two networks are related. It has become increasingly common to collect data consisting of multiple networks, each capturing a different type of relationship between a set of nodes; for example, two online social networks with a shared user base. Many papers have proposed statistical methods that assume that the networks are closely related. Determining whether the networks are related can be critical to the success of these methods. If the networks really are closely related, then these methods often yield more accurate results, but if the networks are not closely related, then these methods can yield highly misleading results.
Fourth-year doctoral student Arash Tarkhan was honored by the Lifetime Data Science section of ASA for his paper, “BigSurvSGD: Big Survival Data Analysis via Stochastic Gradient Descent.”
It is increasingly common to collect and use large heterogeneous datasets in biomedical research. This paper proposes a simple, novel framework to assess the connection between features of a patient and a “time-to-event” outcome (e.g., disease progression or death) in such complex datasets. In particular, it modifies the popular Cox Model to accommodate modern machine learning techniques (e.g., deep learning-based with features such as medical imaging or ECGs). In addition, models in this new framework can efficiently be trained on large amounts of data, using contemporary online optimization algorithms. Tarkhan is currently working with UW Associate Professor of Biostatistics Noah Simon to prepare the manuscript for submission.
Xu (Steven) Wang
Doctoral student Xu (Steven) Wang was also recognized as a winner by the Statistical Learning and Data Science section of ASA for his paper “Statistical Inference for Networks of High-Dimensional Point Processes.”
Motivated by recent applications in neuroscience, the paper developed a statistical inference procedure to identify a network of interactions among multivariate point process data modeled by the Hawkes process. While evaluating the uncertainty of the network estimates is critical in scientific applications, existing methodological and theoretical work have only focused on estimation. To bridge this gap, this paper proposes a high-dimensional statistical inference procedure for multivariate Hawkes process with theoretical guarantees. The paper verifies the theoretical results with extensive simulation and an application to a neuron spike train data set.
Jonathan Fintzi (PhD, ’18)
Recent UW Biostatistics graduate Jonathan Fintzi received the Norman Breslow Award given to the top paper in the Statistics in Epidemiology section of ASA. His paper is titled, “A Linear Noise Approximation for Stochastic Epidemic Models Fit to Partially Observed Incidence Counts.”