Show/Hide University Links

UTSA College of Business
Management Science and Statistics News
Archives

Seminar Series: Spring 2014
The seminar series will usually take place on Fridays in a Business Building room, but the exact time and location could be different due to a variety of factors including room availability in the Business Building. For actual direction to the Business Building, please see campus map. For additional information contact Dr. Kefeng Xu, (210) 458-5388.

Thursday, Jan. 23, 2014 - 2:30-3:30pm, Business Building 4.02.10 (Executive Conference Room)

  • Presenter: Yuli Liang & Chengcheng Hao, Department of Statistics, Stockholm University, Stockholm, Sweden

  • Presentation Title:

1. “A study of multilevel models with block circular symmetric covariance structures” (by Yuli Liang)

  • Abstract: Our work concerns the study of multilevel models with speci c patterned covariance structures and addresses the issues of maximum likelihood estimation. In particu- lar, circular symmetric hierarchical data structures are considered. Models which covariance structures re ect both circularity and exchangeability present in the data can be widely used in di erent applications, with early examples from psychometric and medical research. Two derived patterns of the covariance matrices which characterizes models under consideration. The relationship between these two patterned covariance ma- trices was investigated and it has been veri ed they are similar matrices. New expressions for the eigenvalues of block circular symmetric matrices are obtained which take into account the block structure. Maximum likelihood estimation of balanced multilevel models with block circular symmetric covariance matrices is discussed. We show that explicit maximum likelihood estimators of variance com- ponents exist under certain restrictions on the parameter space.

2. “Local Influence Analysis and Cross-over Studies” (by Chengcheng Hao)

  • Abstract: With a special reference to cross-over design models with random individual effects, the purpose of this dissertation is to develop new methodology to detect influential observations in the context of mixed linear models with explicit maximum likelihood estimators (MLEs). Case-weighted perturbation schemes within and between subjects in mixed models are constructed. It is emphasised that perturbations should be performed under the restriction that explicit MLEs can be obtained in the perturbed model. Two influence functions, the delta-beta influence and variance-ratio influence, are tools to evaluate the influence on the estimates of mean parameters and variance parameters, respectively, with respect to the used perturbations. The proposed approach, named the delta-beta-based local influence approach, derives the expressions of the delta-beta and variance-ratio influences for two specific cross-over designs. In both the AB|BA design (2 X 2 cross-over design) and the ABBA|BAAB design, the applied influence functions turn out to have closed-form expressions of residuals from the unperturbed models. Some graphical tools are also presented.

 

Friday, Feb. 7, 2014, 2-3pm, Business Building, 3.02.16.

  • Presenter: Ted Chang, Professor, University of Virginia

  • Presentation Title: Compositional Data Analysis

  • Abstract: Suppose we have a sample of rocks and x1, ..., xp is the weight of p minerals within the rock. Usually we assume that x1 + . . . + xp represents the total weight of a sample element, so xp might be 'other'. x1, ..., xp are called the 'open' variables. Let yi = xi/(x1 + . . . + xp) denote the proportions, or 'closed' variables; y1 + . . . + yp = 1. This is an example of compositional data. Another example might be data on time allocation: e.g. x1 is the time spent eating, x2 the time spent watching TV, etc.. John Aitchison (JRSS B 1982) proposed an approach to analyzing this type of data. We examine the geometry of moving from open to closed variables and, in that light, the mathematical attractiveness of the Aitchison approach.

  • Speaker Bio: Professor Chang is a world class scholar who has developed many statistical procedures on manifolds, including spherical regression, robust statistics on sphere, statistics on Rigid Body Motion on Sphere and in Euclidean Space, statistics on Stiefel Manifold and Grassmann Manifold, and the reconstruction of the past position of tectonic plates. The common theme of his work is the use of geometric information in statistics.

Friday, Feb. 14, 2014, 2-3pm, Business Building 4.02.10 (Executive Conference Room)

  • Presenter: Peter Westfall, Horn Professor of Statistics, Texas Tech University

  • Presentation Title: Probabilistic Views of Big Data

  • Abstract: Probabilistic thinking has been diminishing in the current era of “big data” and the “data scientist.” In this talk, I explain why this is exactly the wrong response to big data. Rather, probabilistic reasoning is more important than ever because of big data; examples from business intelligence and bioinformatics are given.

 

 

Friday, Feb. 28- 12:15-1:15pm, Business Building 4.02.10 (Executive Conference Room)

  • Presenter: Professor N. Balakrishnan, McMaster University, Hamilton, Ontario, Canada

  • Presentation Title: Likelihood Inference for Left Truncated and Right Censored Data

  • Abstract: In this talk, I will first introduce a power transformers data and motivate the consideration of analysis of left-truncated and right-censored data. I will then present the development of an exact EM-algorithm for this analysis, and then present the results for some specific lifetime distributions of interest. I will also describe the problem of model discrimination based on likelihood criterion and finally present the results of a simulation study and an illustrative example.

Friday, Feb. 28- 2-3pm, Business Building 4.02.10 (Executive Conference Room)

  • Presenter: Mike Daniels, Professor, Section of Integrative Biology, Division of Statistics & Scientific Computation, The University of Texas at Austin

  • Presentation Title: A Flexible Bayesian Approach to Monotone Missing Data in Longitudinal Studies with Informative Missingness with Application to An Acute Schizophrenia Clinical Trial

  • Abstract: We develop a Bayesian nonparametric model for a longitudinal response in the presence of nonignorable missing data. Our general approach is to first specify a {\em working model} that flexibly models the missingness and full outcome processes jointly. We specify a Dirichlet process mixture of missing at random (MAR) models as a prior on the joint distribution of the working model. This aspect of the model governs the fit of the observed data by modeling the observed data distribution as the marginalization over the missing data in the working model. We then separately specify the conditional distribution of the missing data given the observed data and dropout. This approach allows us to identify the distribution of the missing data using identifying restrictions as a starting point. We propose a framework for introducing sensitivity parameters, allowing us to vary the untestable assumptions about the missing data mechanism smoothly. Informative priors on the space of missing data assumptions can be specified to combine inferences under many different assumptions into a final inference and accurately characterize uncertainty. These methods are motivated by, and applied to, data from a clinical trial assessing the efficacy of a new treatment for acute Schizophrenia.

  • Joint work with Antonio Linero at the University of Florida

 

Wednesday, March 5, 2014, 4:00 pm -5:30 pm, Business Building, 3.02.16.

  • Presenter: Bazoumana Kone, Ph.D. Candidate (Dissertation Defense for Ph.D. in Applied Statistics)

  • Presentation Title: BLOCK PREDICTION INTERVALS

  • Abstract: In environmental assessment, such as clean up of contaminated regions (e.g. with dioxin), it is important for scientists (or decision makers) to predict the average amount (called block averages in the geostatistical literature) of contaminant present in the region in order to more effectively remediate the contamination. In geostatistics, block averages are regarded as an integral of random fields over bounded regions. An important problem in predictive inference of block average is the prediction interval of block average value when the distribution of the random fields is not Gaussian. In this dissertation, we propose methods to construct prediction intervals for integrals of Gaussian and non-Gaussian random fields over bounded regions based on observations at a finite set of sampling locations. For Gaussian random fields, we propose two bootstrap calibration algorithms, termed indirect and direct, aimed at improving upon plug-in prediction intervals in terms of coverage probability. We also propose two bootstrap methods for non-Gaussian random fields. The first method relies on the lognormal distributional assumption while the second method does not require any distributional assumptions. Simulation studies are performed to show the effectiveness of all four methods, and these methods are applied to estimate spatial averages of chromium and cadmium traces in a potentially contaminated region in Switzerland.

  • Supervising Professor: Victor De Oliveira, Ph.D.

 

UTSA Homepage | College of Business Homepage | Events | Community | Giving | Directory | Contact Us | Site Map
College of Business — The University of Texas at San Antonio
One UTSA Circle — San Antonio, TX 78249-0631— (210) 458-4313 FAX (210) 458-4308

Please send comments about this site to the College of Business Webmaster
Privacy & Security Policies
TSA Homepage College of Business Homepage