Fall 2022: Hosted by Dr. Xueheng Shi.
Seminars will take place every Wednesday from 3pm-4pm in Hardin Hall 49. You will find the classroom located in the basement of the north tower of Hardin Hall. The seminars are open to all UNL Students, Staff, and Faculty. If you would like to attend, but are not a part of our Statistics department, please fill out the information on the Webform below and you will receive a Zoom link via email.
Wednesday, September 14th
Dr. Xiang Zhu, Department of Statistics and Huck Institutes of the Life Sciences, The Pennsylvania State University
Bayesian regression of genome-wide association summary statistics
Large-scale genome-wide association studies (GWAS) have markedly improved our understanding of how common variation in the human genome affects complex traits and diseases. Regression models have been widely used to analyze GWAS, but existing methods often require input data at the individual level, which are hard to obtain due to many administrative issues. Here we provide a Bayesian framework for multiple regression without the need of individual-level data. Specifically, we derive a "Regression with Summary Statistics" (RSS) likelihood function of the multiple regression coefficients based on the univariate regression summary statistics, which are easily available in GWAS. We combine the RSS likelihood with prior distributions that are specifically designed for a wide range of genetic applications, such as heritability estimation, phenotype prediction, pathway enrichment and gene prioritization. To estimate posterior distributions, we develop efficient Markov chain Monte Carlo and variational inference algorithms that scales well with millions of genetic variants. Applying RSS to a host of real-world GWAS summary statistics, we demonstrate that RSS not only achieves similar performance in settings where existing methods work, but also enables novel analyses and discoveries that existing methods cannot deliver. The software implementing RSS methods is available at https://github.com/stephenslab/rss.
About the Speaker:
Xiang Zhu has been an Assistant Professor of Statistics and Life Sciences at the Pennsylvania State University since 2020, and a Biostatistician (Without Compensation) at U.S. Department of Veterans Affairs Palo Alto Health Care System since 2018. He received his PhD in Statistics from The University of Chicago in 2017, and he was a Stein Fellow at Stanford University in 2017-2020. His research focuses on developing new statistical and computational methodology to mine large-scale and high-throughput genomic data collected from diverse human populations
Wednesday, September 21st
Dean Dustin, PhD Candidate, Department of Statistics at the University of Nebraska-Lincoln. Advised by Dr. Clarke.
Testing for Important Components of Posterior Predictive Variance
We propose a method to decompose the posterior predictive variance using the law of total variance and condition on a finite dimensional discrete random variable. This random variable summarizes various features of modeling that are used for prediction. Later, we test which terms in the decomposition are small enough to ignore. It allows to identify which of the discrete random variables are most important to prediction intervals.
About the Speaker:
Dean Dustin is a PhD Candidate from Manchester, New Hampshire. Dean received his bachelor's degree in Mathematics from Plymouth State University in New Hampshire. At UNL, he has taught classes such as STAT 218, and also worked as a research assistant exploring model uncertainty and prediction. Dean plans to graduate in December 2022 and will be going into the industry to work in quantitative finance.
Wednesday, October 5th
Dr. Susan VanderPlas
Reproducible Science: Statistics, Forensics, and the Law
In science, we strive to design and create reproducible experiments that can be replicated by other researchers. We also usually require that scientists use good experimental design and approved statistical methods for results to be judged credibly. In this talk, I'll review the status quo in forensics: what studies support the use of forensic comparison methods as reliable? Are these studies reliable based on the criteria used in other disciplines? I'll discuss two major, national reports about the validity of forensic science, why statistics as a discipline is critically important to forensics - and what statisticians can do and are doing to improve forensic science.
About the Speaker:
Dr. Susan VanderPlas is an assistant professor in the Statistics Department at the University of Nebraska, Lincoln, researching the perception of statistical charts and graphs, and applying computer vision and machine learning techniques to image data. She also works with the Center for Statistical Applications in Forensic Evidence (CSAFE) at Iowa State University, developing statistical methods for examination of bullets, cartridges, and footwear.