“Breaking Data Science”

Dr. Robert Brunner, University of Illinois at Urbana-Champaign

Date: Thursday, October 22, 2015

Time: 3:30 PM – 4:30 PM

Location: 269 Everitt

Sponsor: Department of Statistics

Abstract:

We live in an increasingly digital world, where an ever-growing quantity of information is generated, collected, and archived. The transformation we are witnessing produces new opportunities across many seemingly unrelated academic fields, leading to the development and growth of a new field: Data Science. In this seminar, I will briefly introduce this new field before reviewing my personal transformation into a Data Scientist. I will conclude with a brief discussion on what I feel the University of Illinois should do to build a data science program that will position us as international leaders in this new field.

http://illinois.edu/calendar/detail/1439?eventId=32978618&calMin=201510&cal=20151019&skinId=13335

“A data-driven approach to conditional screening of high dimensional variables”

Dr. Hyokyoung Hong, Michigan State University

Date: Thursday, October 08, 2015

Time: 3:30 PM – 4:30 PM

Location: 269 Everitt

Sponsor: Department of Statistics

Abstract:

Marginal screening is a widely applied technique to handily reduce the dimensionality of the data when the number of potential features overwhelms the sample size. Due to the nature of the marginal screening procedures, they are also known for their difficulty in identifying the so- called hidden variables that are jointly important but have weak marginal associations with the response variable. Failing to include a hidden variable in the screening stage has two undesirable consequences: (1) important features are missed out in model selection; and (2) biased inference is likely to occur in the subsequent analysis. Motivated by some recent work in conditional screening, we propose a data-driven conditional screening algorithm, which is computationally efficient, enjoys the sure screening property under weaker assumptions on the model, and works robustly in a variety of settings to reduce false negatives of hidden variables. Numerical comparison with alternatives screening procedures are also made to shed light on the relative merit of the proposed method. We illustrate the proposed methodology using a leukemia microarray data example.

http://illinois.edu/calendar/detail/1439?eventId=32932610&calMin=201510&cal=20151005&skinId=13335

Statistics Seminar

“Variance estimation for high-dimensional linear models: Fixed-effects, random-effects, and quadratic forms”

Dr. Lee Dicker, Rutgers University

Date: Thursday, September 24, 2015

Time: 3:30 PM – 4:30 PM

Location: 269 Everitt

Sponsor: Department of Statistics

Abstract:

In regression analysis with repeated measurements, such as longitudinal data and panel data, structured covariance matrices characterized by a small number of parameters have been widely used and play an important role in parameter estimation and statistical inference. To assess the adequacy of a specified covariance structure, one often adopts the classical likelihood-ratio test when the dimension of the repeated measurements (p) is smaller than the sample size (n). However, the assessment becomes challenging when p is bigger than n, since the classical likelihood-ratio test is no longer applicable. This talk will focus an adjusted goodness-of-fit test, which is designed to examine a broad range of covariance structures under the scenario of “large p, small n”. The analytical examples will be presented to illustrate the effectiveness of adjustment for assessing the goodness-of-fit of covariance. In addition, large sample properties of the proposed test are established. Moreover, simulation studies and a real data example are provided to demonstrate the finite sample performance and the practical utility of the test.

http://illinois.edu/calendar/detail/1439?eventId=32825570&calMin=201509&cal=20150921&skinId=13335

Statistics Seminar

“Tests for Covariance Structures with High-dimensional Repeated Measurements”

Dr. Ping-Shou Zhong, Michigan State University

Date: Thursday, September 17, 2015

Time: 3:30 PM – 4:30 PM

Location: 269 Everitt

Sponsor: Department of Statistics

Abstract:

In regression analysis with repeated measurements, such as longitudinal data and panel data, structured covariance matrices characterized by a small number of parameters have been widely used and play an important role in parameter estimation and statistical inference. To assess the adequacy of a specified covariance structure, one often adopts the classical likelihood-ratio test when the dimension of the repeated measurements (p) is smaller than the sample size (n). However, the assessment becomes challenging when p is bigger than n, since the classical likelihood-ratio test is no longer applicable. This talk will focus an adjusted goodness-of-fit test, which is designed to examine a broad range of covariance structures under the scenario of “large p, small n”. The analytical examples will be presented to illustrate the effectiveness of adjustment for assessing the goodness-of-fit of covariance. In addition, large sample properties of the proposed test are established. Moreover, simulation studies and a real data example are provided to demonstrate the finite sample performance and the practical utility of the test.

http://illinois.edu/calendar/detail/1439?eventId=32511447&calMin=201509&cal=20150914&skinId=13335

A talk giving a gentle introduction to Regression Quantiles
will by given by  Steve Portnoy: Wed, Sep 16, 2015 3pm ‐ 4 plus,
Henry Administration Building, Room 138. If there is time, and you
want to see a couple of additional examples, I might suggest loading
the R-package “quantreg” (of Roger Koenker) and bringing your laptop.

I have a couple of undergraduates doing a reading-research course
on inference for regression quantiles with me this fall, and plan
to give a few lectures. Several statistics students (and faculty)
expressed an interest, so I have a moderately large room and welcome any people who might have heard of the topic and wondered what a regression quantile is.

– Steve Portnoy

“Changepoints and Associated Climate Controversies”

Robert Lund, Clemson University

Date: Thursday, September 10, 2015

Time: 3:30 PM – 4:30 PM

Location: 269 Everitt

Sponsor: Department of Statistics

Abstract:

This talk overviews changepoint issues in climatology. Changepoints (inhomogeneities) are ubiquitous features in climatic time series, arising, for example, when stations relocate or instrumentation is changed. Changepoints confound many inference problems and are very important data features. Here, we show why changepoint information is essential in making accurate trend conclusions. Examples are given where inferences are questionable when changepoints are ignored. The talk will delve into two recent contentious climate issues: 1) the recent increase in Atlantic Basin hurricanes; and 2) the “warming hole” (lack of warming) seen in the Eastern United States.

http://illinois.edu/calendar/detail/1439?eventId=32777743&calMin=201509&cal=20150908&skinId=13335

Statistics Seminar

“Overview of a Statistical Genetics and Genomics Research Program”
Dr. Alexander Lipka, University of Illinois, Urbana-Champaign

Date: Thursday, May 07, 2015
Time: 3:30 PM – 4:30 PM
Location: 165 Everitt
Sponsor: Department of Statistics, Dave Zhao

Abstract:
A statistical genetics and genomics research program should focus on developing new statistical approaches for analyzing genotypic and phenotypic data, establish interdisciplinary collaborations that increase our understanding of the genetic and genomic contributions to phenotypic variation, and provide mentoring opportunities for students and postdoctoral researchers. With this in mind, I review how I use my background in statistical genetics to develop statistical approaches for analyzing genotypic and phenotypic associations, and participate in interdisciplinary projects that focus on increasing the nutritional value of maize grain. Some specific topics discussed include the genome association and prediction integrated tool (GAPIT) R package, my role in developing analytical pipelines to identify loci associated with provitamin A and vitamin E content in maize grain, and the development of java plugins that use information criterion-based model selection. I then discuss some specific research topics I would like to investigate in the future, and highlight relevant collaboration opportunities.

http://illinois.edu/calendar/detail/1439?eventId=32183643&calMin=201505&cal=20150501&skinId=13335