All talks will be held in CAS 109 unless otherwise specified.
Spring 2016 Colloquiums
Spring 2016 Masters Paper Presentations
Mon, May 9, 11:00-1135 Lihong Yin
Tue, May 10, 11:00-1135 Kaps Chulu
Thurs, May 12, 11:00-11:35 Yathish Kolli
2:00-2:35 Ryan Brown
2:40-3:15 Ryan Moran
3:15-3:40 Xuerong Liu
Thrs, Mar 1 2:15 - 02:45 Bing Liu
Fri, Apr 8 2:30-3:00 David Arens
Thur, Apr 14 2:00-2:35 Chris Knapp
Tue, May 3 2:30-3:05 Siegfried Anyomi
Wed, May 4 2:40-3:15 Chan Wang
Thur, May 5 2:00 - 02:35 Dirk Bullock
2:40 - 3:15 Dustin Auble
Fri, May 6 2:00-2:35 George Anim
2:40-3:15 Ahlam Al Khodidi
Thursday, Nov 19 4:00-5:00, Dr. William Bush (Case Western Reserve University) "Mixed Models for Estimating Heritable Components of Age-Related Macular Degeneration”
Age-related macular degeneration (AMD) is the leading cause of irreversible visual loss in the elderly in developed countries and typically affects more than 10 % of individuals over age 80. AMD has a large genetic component, with heritability estimated to be between 45% and 70%. Numerous variants have been identified and implicate various molecular mechanisms and pathways for AMD pathogenesis but those variants only explain a portion of AMD's heritability. Using mixed models, we have estimated the cumulative genetic contribution of common variants on AMD risk for multiple pathways related to the etiology of AMD. We have also explored computationally intensive ways of partitioning genetic variance into additive, dominant, and interactive components, and have validated these approaches via simulation studies. To date, we have found that the bulk of variance in AMD risk is attributable to additive effects of the complement system.
Thursday, Nov 5 4:00-5:00, Dr. Yuping Wu (Cleveland State University) "Assessing the performance of prediction models"
One of the most common questions about regression of binary and survival outcomes is “How do I know if my model is a good fit to the data?”. Some traditional measures of overall model performance, such as C statistic and area under the receiver operating characteristic (ROC) curve (AUC), along with several new measures including net reclassification improvement (NRI) and integrated discrimination improvement (IDI) will be presented. Several collaborative projects with the Cleveland Clinic will be illustrated. Graphs of model validation and calibration will be displayed.
Thursday, Oct 22 4:00-5:00, Dr. J. Sunil Rao (University of Miami) "The E-MS Algorithm : Model Selection with Incomplete Data"
We propose a procedure associated with the idea of the E-M algorithm for model selection in the presence of missing data. We will motivate this by demonstrating shortcomings of existing approaches based on the E-M. The new idea extends the concept of parameters to include both the model and the parameters under the model, and thus allows the model to be part of the EM iterations. We develop the procedure, known as the E-MS algorithm, under the assumption that the class of candidate models is finite. Some special cases of the procedure are considered, including E-MS with the generalized information criteria (GIC), and E-MS with the adaptive fence (AF; Jiang, Nguyen and Rao, 2008). We prove numerical convergence of the E-MS algorithm as well as consistency in model selection of the limiting model of the E-MS convergence, for E-MS with GIC and E-MS with AF. We study the impact on model selection of different missing data mechanisms. Furthermore, we carry out extensive simulation studies on the finite-sample performance of the E-MS with comparisons to other procedures. The methodology is also illustrated on a real data analysis involving QTL mapping for an agricultural study on barley grains. This is joint work with Jiming Jiang of UC-Davis and Thuan Nguyen of Oregon Health Sciences University.
Thursday, Oct 1, 4:00-5:00 Dr. Dan Ralescu (University of Cincinnati) “Mixed Models of Uncertainty”
We will discuss several models which take into account a mixture of stochastic and non-stochastic (fuzzy) information. Specifically, we will describe the concept of fuzzy random variables (as an extension of random sets), discuss the expected value, and statistical applications to estimation and testing of hypotheses. Then we will look at the probability of a fuzzy event, and its different definitions, and applications. Finally, if time permits, we will discuss aggregation of fuzzy concepts.
Thursday, April 16th from 2:10-3:00 Dr. Peter Craigmile, (Ohio State University) "Wavelet-based estimation of the long memory parameter in Gaussian non-gappy and gappy time series.
Thursday, March 19th from 2:10-3:00 Dr. Sujay Datta, (University of Akron) "Graphical and Network Models in Bioinformatics".
Thursday, February 12th from 2:10-3:00 Dr. John Tuhao Chen, (Bowling Green State University) "Step-up Confidence Procedures for the Minimum Effective Dose of a Drug."
Tue, Apr 19 12:30-1:30 (CAS438) Dr. Craig L. Zirbel (Dept. of Math & Stats, Bowling Green) Matching RNA motif sequences to known 3D geometries.
Abstract: It is well known that DNA is usually double stranded, with complementary bases (A with T, G with C) making Watson-Crick basepairs and double helices. RNA, by contrast, is usually single stranded, with short segments of complementary bases that fold back on themselves to form short double helices. In between and at the ends of these helices, a variety of interesting geometries can occur, which we call 3D motifs. It is these 3D motifs that make RNA such a versatile actor in the cell; a huge number of important RNAs have been discovered in just the last 10 years. We have full 3D images of many of these 3D motifs from x-ray crystallography, but that is just from a few model organisms. At the same time, we have genomic sequences from many different organisms, and these show sequence variability in the RNA motifs from one organism to another. Thus, an RNA 3D motif is a random system that must maintain a certain geometry but can accommodate some sequence variability. We have made probabilistic models for likely sequence variability in RNA 3D motifs using Stochastic Context-Free Grammars and Markov Random Fields. Many new RNA molecules have been sequenced but we do not yet have experimental results on their 3D geometry. We can use the probabilistic models to match novel sequences to known 3D geometries and thus try to infer the presence of known 3D motifs in new RNA molecules.
Thur, Apr 28 2:00-3:00 Dr. Thaddeus Tarpey (Dept. of Math, Wright State U) Calling Models Wrong for the Wrong Reasons
Abstract: Perhaps the best known quote from statistics is “all models are wrong, some are useful” by George Box (1919-2013). Although useful, this quote can lead one to label a perfectly good model as wrong and hence defective. A model is simply an approximation to the truth and it usually does not make sense to call an approximation wrong. In this talk, examples are provided where models are called wrong due to misconceptions about the meaning of the parameters that define the model. Also, examples are also provided of models that appear useful but can actually lead to incorrect conclusions