Graduate Statistics Seminar : Past Events
Past Seminars
The following seminars have already happened, you may instead view upcoming seminars in this series.
Sept. 24, 2013
Prof. Jing Wang :
3:30 p.m. in SEO 1227
Oct. 1, 2013
Mr. Ting Yuan :
3:30 p.m. in SEO 1227
Oct. 8, 2013
Mr. Ting Yuan :
3:30 p.m. in SEO 1227
Oct. 15, 2013
Mr. Hsin-Hsiung Huang :
3:30 p.m. in SEO 1227
Oct. 22, 2013
Mr. Hsin-Hsiung Huang :
3:30 p.m. in SEO 1227
Oct. 29, 2013
Ms. Zhifan Zhang :
3:30 p.m. in SEO 1227
Nov. 5, 2013
Mr. Yan Sun :
3:30 p.m. in SEO 1227
Nov. 12, 2013
Ms. Ella Revzin :
3:30 p.m. in SEO 1227
Nov. 19, 2013
Ms. Ella Revzin :
3:30 p.m. in SEO 1227
Feb. 11, 2014
Ryan Martin :
3:30 p.m. in SEO 1227
Abstract
This talk will focus on two important principles in statistical inference, namely, sufficiency and conditioning.
After a brief introduction to these two concepts, I will discuss the exaggerated importance of sufficiency
compared to conditioning, and present some new perspectives based on tools from partial differential equations.
I will attempt to explain the general ideas through simple examples.
Feb. 25, 2014
Gib Bassett :
3:30 p.m. in SEO 636
Abstract
In this talk, I will discuss some open problems concerning the recursively generated distribution functions. Questions include the following:
for a given distribution function, what recursively generated sequence has limit close (or equal) to the given one?, and, for a given class of
recursive sequences, what kind of limit distributions are possible? A good introduction to the problem can be found in Sections 5 and 6 of the
paper: Rousseeuw and Bassett, "The Remedian: A Robust Averaging Method for Large Data Sets," <i>Journal of the American Statistical Association</i>, 1990.
March 4, 2014
Ryan Martin :
3:30 p.m. in SEO 1227
Abstract
Bayes's theorem provides the tool for combining prior beliefs with evidence from data, and there are general sufficient conditions
that guarantee the corresponding Bayesian posterior distribution will be consistent, i.e., will concentrate around the true parameter
asymptotically. Non-trivial examples of inconsistent posterior distributions are scarce, so it's not clear if the existing sufficient
conditions are the right things. Towards understanding what it takes for posterior consistency, I will show that by making a minor
adjustment to the formula of Bayes, namely, taking an arbitrary fractional power on the likelihood before combining with the prior, will avoid all
known non-trivial examples of inconsistency. That is, the corresponding pseudo-Bayes posterior based on the fractional likelihood is,
in a certain sense, always consistent. The key question is: how to make use of this interesting phenomenon? I will discuss some possible
answers.
March 11, 2014
Xin Wang :
3:30 p.m. in SEO 636
Abstract
This talk is about the optimal design problem under a regression model with random block effects.
I will give a introduction to the problem setup. Then I will concentrate on results available in three papers that I read.
These include results under an exact design setup and an approximate design setup.
March 18, 2014
Yi Lin :
3:30 p.m. in SEO 636
Abstract
This talk is on experimental design using a Bayesian decision theory
framework. I will give an introduction on how to determine Bayesian optimal
design objectives. Then I will talk about some results in the literature on
the Bayesian optimal criteria for linear and nonlinear design problems.
April 8, 2014
Ken Fujimoto :
3:30 p.m. in SEO 636
Abstract
I will introduce the basic theory and assumptions of the 1-parameter, 2-parameter, and 3-parameter logistic item response (IRT) models and a Markov chain Monte Carlo algorithm used to estimate the parameters of an IRT model. These measurement models are frequently used to analyze test data, such as the data arising from the GRE, MCAT, GMAT, and ACT tests. IRT models establish a mathematical relationship between a person's ability level and the technical properties of an item (e.g., item difficulty, discrimination level, and/or guessing).
April 15, 2014
Ken Fujimoto :
3:30 p.m. in SEO 636
Abstract
Item response theory (IRT) models are useful measurement tools to examine whether an item's difficulty level is the same across subgroups of examinees (e.g., gender and race). This concept has to do with whether an item is "fair" for all subgroups of examinees. I will discuss how to test whether an item is "fair," or whether it displays differential item functioning (DIF) across subgroups of examinees when the subgroup membership is known. I will then discuss extensions to the traditional IRT model that allows for the detection of whether an item displays DIF across latent subgroups of examinees. These IRT extensions incorporate finite-mixture and infinite-mixture modeling.
April 22, 2014
Qingyang Zhang :
3:30 p.m. in SEO 636
Abstract
Traditional cancer studies have been mainly focusing on single
gene or single type of data including mRNA expression, DNA methylation, copy number variation, and etc.
However, such analyses lack power to reveal the molecular mechanisms from the view of system
biology. In this talk, I will present an integrative framework to identify important genetic and epigenetic
features and to quantify the causal relations among these features. I will first talk about what is a Bayesian
Network and what the TCGA data looks like. Then I will introduce the proposed feature selector and Bayesian
Network model. Simulated and real data sets will be used for illustration.
April 29, 2014
Jing Wang :
3:30 p.m. in SEO 636
Abstract
We develop a two-stage adaptive optimal (AO) two-treatment
crossover design with binary endpoint; compared with the popular Internal
Pilot (IP) design, the AO design reduces the sample size as well as
maintains the power of hypothesis testing.
Feb. 2, 2016
Raymond Mess :
4 p.m. in SEO 636
Abstract
I will talk a little bit about possible applications of inferential model theory to the field of statistical graphical models. The talk will be informal and introductory.
Feb. 9, 2016
Tian Tian :
4 p.m. in SEO 636
Abstract
The use of discrete choice experiments (DCEs) for modeling real marketplace choices, in both fundamental and applied research, has gained much attention recently. To construct proper choice designs, one needs the help of efficient
search algorithms, among which the coordinate-exchange algorithm (CEA) has shown itself to
work very well under the widely used multinomial logit discrete choice model. However, due to
the discrete nature of choice designs, there are no computationally feasible ways to verify
the real performance of the resulting designs. In this talk, I'll discuss an approach of evaluating the
performance of the CEA.
Empirical study was conducted to show that the CEA is indeed highly efficient for deriving homogeneous optimal designs.
Feb. 16, 2016
Nick Syring :
4 p.m. in SEO 636
Abstract
I will introduce the problem of identifying boundaries in images observed with random noise. I will present a Gibbs model solution, which combines elements of machine learning and Bayesian statistics. I have produced a proof that the proposed model converges at the minimax rate, and I show through simulations the competitive performance of the Gibbs model. If there is sufficient interest, I may share the details of the proof at a later date.
Feb. 23, 2016
Jennifer Pajda-De La O :
4 p.m. in SEO 636
Abstract
Probability measures and mild convergence.
March 8, 2016
Jennifer Pajda-De La O :
4 p.m. in SEO 636
Abstract
Cluster analysis is one common tool in data mining. I will give a brief introduction to cluster analysis and why it is important. I will then review the k-Means clustering algorithm and give an example of how to apply this method.
March 15, 2016
Keyu Nie :
4 p.m. in SEO 636
April 5, 2016
Statistics Grad Students :
4 p.m. in SEO 427
April 12, 2016
Statistics Grad Students :
4 p.m. in SEO 636
April 26, 2016
Yi Lin :
4 p.m. in SEO 636
Sept. 7, 2016
Graduate students from stat program :
3 p.m. in SEO 512
Abstract
Six students who did/are doing internships at various companies/agencies are
invited to talk about their experiences.
Sept. 14, 2016
Xianwei Bu :
3 p.m. in SEO 512
Abstract
Multinomial logistic models have been widely used for categorical responses or multinomial responses.
In order to construct a general framework towards D-optimal designs for these models, we unify all the
four models into a common form and extend them to fit different model assumptions. We explore the design
space of these models, derive a simplified form of the Fisher information matrix, and obtain explicit
formulas of its determinant. We also develop efficient numerical algorithms for searching D-optimal designs.
Sept. 21, 2016
Xindi Wang :
3 p.m. in SEO 512
Sept. 28, 2016
Tian Tian :
3 p.m. in SEO 512
Abstract
To see some basic optimal design concepts from an intuitive point of view...
Oct. 12, 2016
Raymond Mess :
3 p.m. in SEO 512
Abstract
Raymond will talk about some bio-statistics and clinical trials related courses at the west campus.
Oct. 19, 2016
Shuai Hao :
2 p.m. in SEO 512
Abstract
Oral Practice for Shuai Hao.
Oct. 26, 2016
Yi Hua :
3 p.m. in SEO 512
Abstract
Oral practice for Yi Hua
Nov. 2, 2016
Maryam Emami Neyestanak :
3 p.m. in SEO 512
Abstract
As part of Concentrating Solar Power Capital Cost and Expected O&M project, we model the direct solar irradiance using stochastic learning methods, clear sky data and the NWS database. Project partners NREL, Northwestern University, Colorado School of Mines, Argonne National Laboratory, and SolarReserve develop detailed performance and cost models. We use ARMA models with additional predictors from the National Digital Forecast Database (NDFD) of the National Weather Service (NWS), and our approach is motivated in large part by the work of Marquez and Coimbra, although we employ simpler forecasting models than their neural networks.
Nov. 9, 2016
Tian Tian :
3 p.m. in SEO 512
Abstract
I'll talk briefly about optimal design theory in dose-finding problems with late-onset toxicities. Models we built and methods we applied. Algorithm structure will also be illustrated.
It's a poster competition practice for ICSA Midwest Meeting.
Nov. 30, 2016
Jennifer, Tian, Raymond :
3 p.m. in SEO 512
Abstract
Three students will talk briefly about official job interviews, dressing code, preparation, presentation, etc.
Feb. 14, 2018
Jie Yang :
3 p.m. in SEO 636
Abstract
In this talk, I will briefly introduce several active research areas of mine, including related publications, current research projects, and potential research topics for students. In the area of optimal design theory, we developed some theoretical results and numerical algorithms for D-optimal designs under generalized linear models and multinomial logistic models. We expect to extend our results to more general multivariate generalized linear models and big data analysis. In the area of bioinformatics, we developed a numerical vectorization approach for classifying viral genomes and protein sequences. We also proposed a high-dimensional classification approach based on a permamental process. We are working on the permanental approach for big data analysis and microbiome data analysis. In the area of financial mathematics, we have proposed two nonparametric approaches for estimating the risk-neutral densities using market option prices. We are working on the consistency of the proposed approaches.
April 24, 2019
Hy Tran :
3 p.m. in 636 SEO
Abstract
I will be discussing the role of a statistician in applied research.
In particular I will be using examples from my own work.
This talk should be accessible to all statistics students.
Jan. 29, 2020
Yushen Dong :
3 p.m. in 712 SEO
Abstract
Variable selection has been well studied in the recent literatures due to the surge of enormous high dimensional data. Interaction between predictors is commonly expected to exist in all kinds of real applications. Recently some parametric interaction selection methods have been proposed. In this talk, we will present a new method to perform nonparametric interaction selection and screening, based on the measurement error selection likelihood approach. This method uses local constant smoothing and backfitting algorithm to perform main and interaction selection for additive model. Resulting solution path will exhibit the importance of predictors. Finite-sample simulation shows this method performs well.
Feb. 5, 2020
Dani Tucker :
3 p.m. in 712 SEO
March 18, 2020
Raymond Mess :
3 p.m. in 636 SEO
April 1, 2020
Xuelong Wang :
3 p.m. in Zoom: https://uic.zoom.us/j/739425989
April 8, 2020
Yichao Wu :
3 p.m. in Zoom Meeting
April 15, 2020
Maryam Emami Neyestanak :
4 p.m. in Zoom
April 22, 2020
Keren Li :
3 p.m. in Zoom
Abstract
We propose a fast and efficient strategy, called the representative approach,
for big data analysis with linear models and generalized linear models. With a
given partition of big dataset, this approach constructs a representative data point
for each data block and fits the target model using the representative dataset. In
terms of time complexity, it is as fast as the subsampling approaches in the literature. As for efficiency, its accuracy in estimating parameters is better than the
divide-and-conquer method. With comprehensive simulation studies and theoretical justifications, we recommend two representative approaches. For linear models
or generalized linear models with a flat inverse link function and moderate coefficients of continuous variables, we recommend mean representatives (MR). For other
cases, we recommend score-matching representatives (SMR). As an illustrative application to the Airline on-time performance data, MR and SMR are as good as the
full data estimate when available. Furthermore, the proposed representative strategy is ideal for analyzing massive data dispersed over a network of interconnected
computers
April 29, 2020
Ping-Shou Zhong :
4 p.m. in Zoom
Abstract
In this talk, I will share some statistical inference problems for high dimensional and functional data in applications such as genomics and genetics, and neuroimaging. The topics include problems in statistical inference for high dimensional parameters in data with sparse and weak signals, covariance structure model selection, and change-points detection and identification in high dimensional longitudinal/functional data.
Aug. 26, 2020
Statistics Graduate Student Council :
5 p.m. in Zoom
Nov. 8, 2023
Youhan Lu :
3 p.m. in Zoom
Nov. 15, 2023
Yifei Huang :
3 p.m. in 612 SEO