Seminar Schedule


   

2/16/2018

2:30pm 

MCSC 310

 

Collective Nonparametric Density and Spectral Density Estimation with Applications in Bioinformatics

Dr. Mehdi Maadooliat, Assistant Professor

Department of Mathematics, Statistics and Computer Science, Marquette University


Abstract:

In this talk, I review a nonparametric method for collective estimation of multiple bivariate density functions for a collection of populations of protein backbone angles. This collective density estimation approach is widely applicable when there is a need to estimate multiple density functions from different populations with common features. In the second part of the talk, I present an extension of this approach for the simultaneous estimation of spectral density functions (SDFs) for a collection of stationary time series that share some common features. A collective estimation approach pools information and borrows strength across the SDFs to achieve better estimation efficiency. Also, each estimated spectral density has a concise representation using the coefficients of the basis expansion, and these coefficients can be used for visualization, clustering, and classification purposes. The Whittle pseudo-maximum likelihood approach is used to fit the model, and an alternating blockwise Newton-type algorithm is developed for the computation. A web-based shiny App found at “https://ncsde.shinyapps.io/NCSDE” is developed for visualization, training and learning the SDFs collectively using the proposed technique. Finally, we apply our method to cluster similar brain signals recorded by the electroencephalogram for identifying synchronized brain regions according to their spectral densities.

   

3/2/2018

2:30pm 

MSCS 310

 

Many Dimensions and Many Correlations: Statistical Challenges in the Analysis of Modern Biological Data

Dr. Pratyaydipta Rudra

Postdoctoral Scholar, University of Colorado

The increase in the volume of data from biological sciences and the need for integrating several data types to answer complex research questions pose new statistical and computational challenges. These data are not only large, but also complex due to the interaction of many scientific factors. In particular, the data are often multi-dimensional and involve unusual correlation structures among the variables of interest. It is important to develop novel statistical models to quantify the correlations in an optimal way. The talk will cover several strategies to quantify and model correlation in high-dimensional biological data arising from real life problems. 

   

4/6/2018

2:30pm 

MSCS 310

Safety Training: Ergonomics in the Workplace

Greg Hogan from Environmental Health and Safety 

   

4/13/2018

2:30pm 

MSCS 310

Mu Sigma Rho and Learning

Dr. Adam Molnar

Part 1: Inductions into Mu Sigma Rho, the Statistics Honor Society

Part 2: Theories of Learning

Grades and honors such as Mu Sigma Rho reward something we call "learning". Statisticians tend to evaluate learning based on results, such as remembering definitions, solving problems, and writing reports. Philosophers and education professors, on the other hand, often write about theories and frameworks of learning. A researcher's choice of theory will affect investigations conducted and research published by that researcher. This talk will briefly summarize popular theories of learning in mathematics and statistics education, and how each theory affects works published by authors of that theory.

   

4/20/2018

2:30pm 

MSCS 310

Bayesian Analysis for Sparse Functional Data

Sean Ye

PhD Candidate

Abstract: This dissertation mainly presents a novel Bayesian method for sparse functional data. Specifically, two models are proposed, one of which models all individual functions with a common smoothness and the other groups individual functions with heterogeneous smoothness. The proposed method utilizes the mixed effects model representation of the penalized splines for both the mean function and the individual functions. Given noninformative or weakly informative priors, the Bayesian inference on the proposed models are developed and computations are done by using Markov Chain Monte Carlo methods. It has been shown that the proposed Bayesian methods performs well on irregularly spaced sparse functional data, where a traditional mixed effects model may often fail. This dissertation also includes a small section on orthogonal series functional estimation for density functions. 

   

4/27/2018

2:30pm 

MSCS 310 

Identify risk factors associated with HCV mono-infection, HIV mono-infection and HIVHCV co-infection in the southwest of China

Biting Zhou

Acquired Immune Deficiency Syndrome (AIDS) and Hepatitis C are global health problems of unprecedented scale. Moreover, many reports indicate that Hepatitis C virus (HCV) has a higher infection rate among people who lives with Human immunodeficiency virus (HIV). Thus there raises a need to better understanding of risk factors associated with each infection and their co-infection. The aim of this report is to identify risk factors for HCV mono-infection, HIV mono-infection and HIVHCV co-infection among population in the southwest of China, where it is an AIDS epidemic area and also one of the largest areas inhabited by the Yi people. A data set of 3576 enrollees in the Methadone Maintenance Treatment Clinic (MMT) between 2004 and 2012 was utilized. We used univariate associate test and multiple logistic regression model to identify the risk factors for each type of infection. The results indicate that in the studied area the HCV mono-infection rate is associated with gender, age, methods of drug abusing for 6 months, drug injection, frequency of drug usage per day, occupation, and HIV infection status; HIV mono-infection rate is associated with gender, age of initial drug usage, duration of drug usage, drug injection, education, occupation, and HCV infection status; The HIVHCV co-infection is highly associated with drug injection.  Knowing the potential risks in turn helps design more effective risk mitigation strategies to reduce the risk of infection in the area under the study.

*****************

A Comparison:  The Two-Sample Kolmogorov- Smirnov, Cramer-Von Mises, and Anderson-Darling Tests

Jana Alford

The objective of this paper is to compare the two-sample Kolmogorov-Smirnov, Cramer-Von Mises, and Anderson Darling test statistics. We wish to make conclusions on the performance of each test statistic under varying conditions, so that researchers can choose which test to use based on these specific conditions. Random samples following three distributions were examined with different shape, location, and scale parameters. We first determined the minimum sample size needed to achieve a predetermined Type I error rate for each test statistic. We then compared the power of each test statistic at a common sample size to determine which will have the greatest power.

*******************

A Simulation Study on Coherence Estimation of Time Series 

Youngjun Chae 

This report investigates the estimation of spectral coherence between two time series via simulation studies. The coherence estimation depends on both the univariate and bivariate spectral density estimation. The purpose of the study is to compare different kernel smoothing methods for correlated autoregressive time series. The coherence estimation is important for studies in climate change, when the temporal relationship between climate variables is of interest. An example using Oklahoma observatory data is illustrated.