Statistics and Data Science Seminars

COSAM Departments Mathematics & Statistics Research Seminars Statistics and Data Science

Upcoming Statistics and Data Science Seminars

Past Statistics and Data Science Seminars

DMS Statistics and Data Science Seminar
Apr 17, 2024 02:00 PM
354 Parker Hall

ShujieMa

Speaker: Dr. Shujie Ma (University of California at Riverside)

Title: Causal Inference on Quantile Dose-response Functions via Local ReLU Least Squares Weighting

Abstract: In this talk, I will introduce a novel local ReLU network least squares weighting method to estimate quantile dose-response functions in observational studies. Unlike the conventional inverse propensity weighting (IPW) method, we estimate the weighting function involved in the treatment effect estimator directly through local ReLU least squares optimization. The proposed method takes advantage of ReLU networks applied for the baseline covariates with increasing dimension to alleviate the dimensionality problem while retaining flexibility and local kernel smoothing for the continuous treatment to precisely estimate the quantile dose-response function and prepare for statistical inference. Our method enjoys computational convenience, scalability, and flexibility. It also improves robustness and numerical stability compared to the conventional IPW method. We show that the ReLU networks can break the notorious `curse of dimensionality' when the weighting function belongs to a newly introduced smoothness class. We also establish the convergence rate for the ReLU network estimator and the asymptotic normality of the proposed estimator for the quantile dose-response function. We further propose a multiplier bootstrap method to construct confidence bands for quantile dose-response functions. The finite sample performance of our proposed method is illustrated through simulations and a real data application.

DMS Statistics and Data Science Seminar
Apr 10, 2024 02:00 PM
354 Parker Hall

Speaker: Dr. Ted Westling (Assistant Professor, University of Massachusetts Amherst)

Title: Consistency of the bootstrap for asymptotically linear estimators based on machine learning

Abstract: The bootstrap is a popular method of constructing confidence intervals due to its ease of use and broad applicability. Theoretical properties of bootstrap procedures have been established in a variety of settings. However, there is limited theoretical research on the use of the bootstrap in the context of estimation of a differentiable functional in a nonparametric or semiparametric model when nuisance functions are estimated using machine learning. In this article, we provide general conditions for consistency of the bootstrap in such scenarios. Our results cover a range of estimator constructions, nuisance estimation methods, bootstrap sampling distributions, and bootstrap confidence interval types. We provide refined results for the empirical bootstrap and smoothed bootstraps, and for one-step estimators, plug-in estimators, empirical mean plug-in estimators, and estimating equations-based estimators. We illustrate the use of our general results by demonstrating the asymptotic validity of bootstrap confidence intervals for the average density value and G-computed conditional mean parameters, and compare their performance in finite samples using numerical studies. Throughout, we emphasize whether and how the bootstrap can produce asymptotically valid confidence intervals when standard methods fail to do so.

This is joint work with UMass Amherst Statistics PhD student Zhou Tang. A preprint of the paper is available here: https://arxiv.org/abs/2404.03064.

DMS Statistics and Data Science Seminar
Apr 03, 2024 02:00 PM
ZOOM

Speaker: Dr. Panpan Zhang (Assistant Professor, Vanderbilt University Medical Center)

Title: Challenges and Opportunities for Longitudinal Analysis of Neurodegenerative Disorders

Abstract: Alzheimer's disease (AD) and Parkinson's disease (PD) are chronic neurodegenerative disorders that gradually destroy memory, thinking skills, and mobility, causing significant impacts on life quality and economic burden. Longitudinal analysis is a promising tool that helps clinicians and neuroscientists better understand changes in the characteristics of the target population over the continuum of AD (or PD) progression. However, the lengthy course of development of such diseases poses many challenges in biostatistical studies. In this presentation, I will introduce two recent projects respectively focusing on missing covariate problems and mismatching time scale problems arising from the longitudinal modeling of AD and PD. I will showcase the novelty of the proposed methods, but also discuss their limitations and potential improvements. The applications in these two projects are primarily based on the open data from the Parkinson's Progression Markers Initiative (PPMI) and the Alzheimer's Disease Neuroimaging Initiative (ADNI).

DMS Statistics and Data Science Seminar
Mar 20, 2024 02:00 PM
354 Parker Hall

Speaker: Dr. Linbo Wang (University of Toronto)

Title: Sparse Causal Learning

Abstract: In many observational studies, researchers are interested in studying the effects of multiple exposures on the same outcome. Unmeasured confounding is a key challenge in these studies as it may bias the causal effect estimate. To mitigate the confounding bias, we introduce a novel device, called the synthetic instrument, to leverage the information contained in multiple exposures for causal effect identification and estimation. We show that under linear structural equation models, the problem of causal effect estimation can be formulated as an \(\ell_0\)-penalization problem, and hence can be solved efficiently using off-the-shelf software. Simulations show that our approach outperforms state-of-art methods in both low-dimensional and high-dimensional settings. We further illustrate our method using a mouse obesity dataset.

Bio: Linbo Wang is an assistant professor in the Department of Statistical Sciences and the Department of Computer and Mathematical Sciences, University of Toronto. He is also a faculty affiliate at the Vector Institute, a CANSSI Ontario STAGE program mentor, and an Affiliate Assistant Professor in the Department of Statistics, University of Washington, and Department of Computer Science, University of Toronto. Prior to these roles, he was a postdoc at Harvard T.H. Chan School of Public Health. He obtained his Ph.D. from the University of Washington. His research interest is centered around causality and its interaction with statistics and machine learning.

DMS Statistics and Data Science Seminar
Mar 13, 2024 02:00 PM
354 Parker Hall

Speaker: Dr. Sathyanarayanan Aakur, Assistant Professor from Auburn CSSE.

Title: Towards Multimodal Open World Event Understanding with Neuro Symbolic Reasoning.

Abstract: Deep learning models for multimodal understanding have taken great strides in tasks such as event recognition, segmentation, and localization. However, there appears to be an implicit closed world assumption in these approaches; i.e., they assume that all observed data is composed of a static, known set of objects (nouns), actions (verbs), and activities (noun+verb combination) that are in 1:1 correspondence with the vocabulary from the training data. One must account for every eventuality when training these systems to ensure their performance in real-world environments. In this talk, I will present our recent efforts to build open-world understanding models that leverage the general-purpose knowledge embedded in large-scale knowledge bases for providing supervision using a neuro-symbolic framework based on Grenander’s Pattern Theory formalism. Then I will talk about how this framework can be extended to abductive reasoning for natural language inference and commonsense reasoning for visual understanding. Finally, I will briefly present some results from the bottom-up neural side of open-world event perception that helps navigate clutter and provides cues for the abductive reasoning frameworks.

DMS Statistics and Data Science Seminar
Nov 15, 2023 02:00 PM
354 Parker Hall

Speaker: Dr. Raghu Pasupathy (Purdue University)

Title: Batching as an Uncertainty Quantification Device

Abstract: Consider the context of a statistician, simulationist, or an optimizer seeking to assess the quality of \(\theta_n\), an estimator of an unknown object \(\theta \in \mathbb{R}^d\), constructed using data \((Y_1, Y_2,\ldots, Y_n)\) gathered from a source such as a dataset, a simulation, or an optimization routine. The unknown object \(\theta\) is assumed to be a statistical function of the probability measure that generates the stationary time series \((Y_1, Y_2,\ldots, Y_n)\). In such contexts, resampling methods such as the bootstrap or subsampling have been the classical answer to the question of how to approximate the sampling distribution of the error \(\theta_n - \theta\). In this talk, we propose a simple alternative called batching. Batching works by appropriately grouping the data \((Y_1, Y_2,\ldots, Y_n)\) into contiguous and possibly overlapping batches, each of which is then used to construct an estimate of \(\theta\). These batch estimates, along with the original estimate \(\theta_n\), are then combined and scaled appropriately to approximate any functional such as the bias, or the mean-squared error of the error \(\theta_n - \theta\), or to construct the \((1-\alpha)\)-confidence region on \(\theta\). We show that batching, like bootstrapping, enjoys strong consistency and high-order accuracy properties. Furthermore, we show that the weak asymptotics of batched studentized statistics are not necessarily normal but characterizable. In particular, using large overlapping batches when constructing confidence regions delivers consistently favorable performance. A number of theoretical and practical questions about batching are open.

DMS Statistics and Data Science Seminar
Nov 08, 2023 02:00 PM
ZOOM

Speaker: Dr. HaiYing Wang (University of Connecticut)

Title: Rare Events Data and Maximum Sampled Conditional Likelihood

Abstract: We show that the available information about unknown parameters in rare events data is only tied to the relatively small number of cases, which justifies the usage of negative sampling. However, if the negative instances are subsampled to the same level of the positive cases, there is information loss. We derive an optimal sampling probability for the inverse probability weighted (IPW) estimator to minimize the information loss. We further we propose a likelihood-based estimator to further improve the estimation efficiency, and show that the improved estimator has the smallest asymptotic variance among a large class of estimators. It is also more robust to pilot misspecification. The likelihood-based estimator is also generalized to a class of models beyond binary response models. We validate our approach on simulated data, the MNIST data, and a real click-through rate dataset with more than 0.3 trillion instances.

DMS Statistics and Data Science Seminar
Nov 01, 2023 02:00 PM
354 Parker Hall

Dr. Subrata Kundu (George Washington University)

Title: The Statistical Face of a Region under Monsoon Rainfall in Eastern India

(Joint work with Kaushik Jana, Ahmedabad University & Debasis Sengupta, Indian Statistical Institute, Kolkata)

Abstract: A region under rainfall is a contiguous spatial area receiving positive precipitation at a particular time. The probabilistic behavior of such a region is an issue of interest in meteorological studies. A region under rainfall can be viewed as a shape object of a special kind, where scale and rotational invariance are not necessarily desirable attributes of a mathematical representation. For modeling variation in objects of this type, we propose an approximation of the boundary that can be represented as a real-valued function, and arrive at further approximation through functional principal component analysis, after suitable adjustment for asymmetry and incompleteness in the data. The analysis of an open-access satellite data set on monsoon precipitation over Eastern India leads to an explanation of most of the variation in shapes of the regions under rainfall through a handful of interpretable functions that can be further approximated parametrically. The most important aspect of shape is found to be the size followed by contraction/elongation, mostly along two pairs of orthogonal axes. The different modes of variation are remarkably stable across calendar years and across different thresholds for the minimum size of the region.

DMS Statistics and Data Science Seminar
Oct 25, 2023 02:00 PM
354 Parker Hall / ZOOM

Speaker: Dr. Yanyuan Ma (Penn State University)

Title: Doubly Flexible Estimation under Label Shift

Abstract: In studies ranging from clinical medicine to policy research, complete data are usually available from a population P, but the quantity of interest is often sought for a related but different population Q which only has partial data. In this paper, we consider the setting that both outcome Y and covariate X are available from P whereas only X is available from Q, under the so-called label shift assumption, i.e., the conditional distribution of X given Y remains the same across the two populations. To estimate the parameter of interest in population Q via leveraging the information from population P, the following three ingredients are essential: (a) the common conditional distribution of X given Y, (b) the regression model of Y given X in population P, and (c) the density ratio of the outcome Y between the two populations. We propose an estimation procedure that only needs some standard nonparametric regression technique to approximate the conditional expectations with respect to (a), while by no means needs an estimate or model for (b) or (c); i.e., doubly flexible to the possible model misspecifications of both (b) and (c). This is conceptually different from the well-known doubly robust estimation in that, double robustness allows at most one model to be misspecified whereas our proposal here can allow both (b) and (c) to be misspecified. This is of particular interest in our setting because estimating (c) is difficult, if not impossible, by virtue of the absence of the Y-data in population Q. Furthermore, even though the estimation of (b) is sometimes off-the-shelf, it can face the curse of dimensionality or computational challenges. We develop the large sample theory for the proposed estimator and examine its finite-sample performance through simulation studies as well as an application to the MIMIC-III database.

DMS Statistics and Data Science Seminar
Oct 18, 2023 02:00 PM
ZOOM

Speaker: Dr. Francesca Chiaromonte (Penn State University)

Title: Functional Data and its use in Biomedical Applications: Power and Perils.

Abstract: Among the deluge of data that surrounds us, some can be – naturally and rather usefully – represented as curves unfolding over a continuous domain, say time. Think of children’s growth curves, or of the curves depicting deaths or cases during an epidemic. This type of data, called functional data, has shapes – and statisticians have developed powerful approaches to leverage the rich information it conveys. The talk will start with an informal introduction to functional data and some of the techniques that allow us to analyze it. Next, we will illustrate how functional data analysis can support biomedical research with applications concerning childhood obesity and the spread of COVID-19 in Italy. We will conclude with some words of caution on the perils associated with the use of sophisticated and powerful methodology, including functional data analysis.

More Events...

College of Sciences and Mathematics Homepage

Statistics and Data Science Seminars

In the News