DESCRIPTIONS OF CURRENT RESEARCH - 2007/2008

Berliner

My general interest is the implementation of Bayesian analysis in complex settings, with particular attention to geophysical problems. The Bayesian paradigm provides opportunities for the combination of physical reasoning and observational data in a coherent analysis framework, but in a fashion that manages the uncertainties in both information sources. A key to the modeling is the hierarchical viewpoint in which separate statistical models are developed for the physical variables studied and the observations conditional on those variables. Modeling physical variables in this way enables incorporation of scientific models across a spectrum of levels of intensity ranging from qualitative use of physical reasoning to strong reliance on numerical models. Modeling and computational methods are being developed and applied to problems of assessing climate change and its impacts, weather forecasting, glacial dynamics, and medium-range climate prediction.

Browne

I am currently working with Professor John Nesselroade of the University of Virginia on the representation of psychological processes using dynamic factor analysis models. Other current research of mine investigates why commonly used likelihood-ratio-based fit indices for linear latent curve models can improve substantially when error of measurement is increased. This work is being carried out in collaboration with Robert MacCallum of The University of North Carolina.

Calder

My primary research interests are in spatial and spatio-temporal statistics, especially dimension reduction and multivariate methods in both of these areas, and in Bayesian methods and computation. Currently, one of my key research areas is the development of methods for analyzing categorical spatial and space-time data. In addition, I am working with Dr. Noel Cressie (Statistics) and Ph.D. student Hongfei Li (Statistics) on model-based methods for testing for spatial dependence in processes on a lattice and with Dr. Ningchuan Xiao (Geography) on projects related to the visualization of uncertainty in spatial information. I am also involved in several collaborative projects that use Bayesian hierarchical modeling in the following areas of application: (1) exploring the pathways linking concentrations of pollutants in environmental media to concentrations in human biomarkers, with Drs. Peter Craigmile (Statistics), Noel Cressie (Statistics), and Tom Santner (Statistics), and several researchers from Battelle; determining the impact of segregation on the spatial patterning of crime across major U.S. cities, with Drs. Chris Browning (Sociology), Laurie Krivo (Sociology), Mei-Po Kwan (Geography), Ruth Peterson (Sociology); and assessing the impact of fire on the atmospheric concentration of black carbon aerosols in Southeast Asia, with Drs. Darla Munroe (Geography), Tao Shi (Statistics), and Ningchuan Xiao (Geography).

Craigmile

My research interests include time series analysis, longitudinal methods, and spatial statistics. I am interested in the use of spectral and wavelet methods to investigate dependency structures and to analyze periodicities and trends. One application of this is to the study of long memory processes. In the area of spatial statistics, I conduct research in the study of space-time processes and the prediction of spatial exceedances. I enjoy application-oriented research.

Cressie

My main research area is in spatial statistics and spatio-temporal statistics. The technical results in these areas are used as components of hierarchical statistical models for problems in environmental science and engineering. This leads to the development of Bayesian and empirical Bayesian methodology in complex, non-linear systems. Examples of problems I have worked on include long-lead forecasting of the El Nino phenomenon, remote sensing of global environmental processes, ice-stream dynamics, Bayesian statistical exposure modeling from sources to biomarkers, and mapping of disease risk over contiguous small areas.

Critchlow

I am interested in some problems involving phylogenetic tree methods (for example, tree distances), and their connection with well-studied permutation methods.

Dean

My main area of research concerns the design of experiments for large numbers of factors and, in particular, screening to find those factors that most influence a response. A long-term collaboration with Sue Lewis of Southampton University in England is continuing in this area. We have recently edited a volume of invited papers dealing with many and varied aspects of screening. Together with PhD student Danel Dragulic, we are investigating which of several different designs and analysis methods performs the best for different types of screening tasks. Ph.D. student Hyejung Moon (who is being jointly supervised by Thomas Santner and me) is investigating the use of group screening in computer experiments. I am also involved in developing the theory of construction of three different types of factorial experiments arranged in split-plot and split-lot designs, one project with recently graduated Ph.D. student Tena Katsaounis, one with Derek Bingham and Pritam Ranjan, and the third with Sue Lewis and Rob Stapleton.

In the human decision area, I am working with recently graduated Ph.D. student Qing Liu and with Greg Allenby, Professor of Marketing, in studying the effect on product ratings of various numbers of options provided to subjects in a conjoint experiment. A second project, jointly with Steven MacEachern and recently graduated Ph.D. student Shiling Ruan involves the use of the Poisson race model for modelling human decisions. These research topics form part a large NSF-funded project on hierarchical Bayesian methods in consumer behavior.

Goel

My research interests include theoretical and practical aspects of Bayesian hierarchical modeling for nonlinear dynamic systems, longitudinal studies, Bayesian networks, information theory and comparison of experiments, and combining information and data fusion. Currently, my effort is focused on (i) statistical information theory, in collaboration with Professor Ginebra, Polytechnic University of Barcelona, (ii) statistical image processing, pattern recognition and transportation networks modeling and estimation, in collaboration with faculty in Transportation Engineering and Remote Sensing, (iii) Bayesian estimation for non-linear dynamic systems and data-rectification with application in process engineering, in collaboration with Professor Bakshi, Chemical Engineering and (iv) Modeling of Bone Mass Accretion in Young Females based on a 7-years longitudinal study, in collaboration with Dr. Matkovic, Physical Medicine. Statistical information theory studies include foundational issues in characterizing information measures in decision analysis framework. The Transportation engineering applications include developing space-time models for combining information on traffic counts from remote-sensed satellite and airborne sensors and ground data. Remote-sensed data provides spatial images at an instant in time, while the ground-based Automatic Traffic Recorder provides short-term, temporal data at a few fixed locations. The goal is to reduce costs and/or errors in estimating transportation planning parameters. The study also involves image processing for automatic pattern recognition of vehicles in airborne and high-resolution satellite images. Current research funding include, grants from DoT/BTS, and NIH and DOT funded, Ohio State led consortium for Application of Remote Sensing in Transportation Flows.

Hans

My research interests are generally focused on Bayesian methodology and associated computation. Among my main areas of current research are problems of variable selection and model uncertainty in contexts of regression, prediction and complex multivariate modeling with many variables. Relevant modeling and computational issues include specification of prior distributions for model spaces and stochastic search and MCMC methods for exploring model spaces with collinear covariates. Additionally I work on developing parallel computing methods for high dimensional models.

Herbei

I am interested in Bayesian methodology and its applications, with emphasis on atmospheric sciences. Currently I am working with Dr. Kate Calder (Statistics), Dr. Heather Allen (Chemistry) and Ph.D. Student Chris Beekman (Chemistry) on the inverse problem of recovering vertical profiles of trace gases and aerosols from raw spectra. I am also analyzing limnological data in collaboration with Dr. Berry Lyons (Byrd Polar Research Center). On the theoretical side, I am studying rates of convergence for hybrid samplers in ill-posed inverse problems.

Hsu

My research is in the area of multiple comparisons. Currently the emphasis is on using the partitioning principle to control the generalized familywise error rate. One of my projects involves conducting proof-of-concept experiments to demonstrate the importance of proper statistical design and analysis in microarray experiments toward pharmacogenomics. Another project involves finding hidden conditions for re-sampling methods such as permutation tests to control multiple testing error rates.

Kaizar

I have several research interests that are tied together by the common theme of using all available data to answer questions, which are primarily focused in the health and social sciences. My current research is about appropriately combining data collected from randomized controlled trials with data that populates administrative databases in unified analyses to reduce or eliminate the potential selection biases inherently introduced by both data collection methods. To accomplish this, I use Bayesian hierarchical models. I am interested in sensitivity analyses and appropriate prior specification for these models. I also work with data collected via sample surveys.

Kubatko

My research interests are in statistical genetics, particularly the estimation of phylogenetic trees from nucleotide sequence data. My recent work in this area has focused on modeling the relationship between phylogenies based on individual genes and those representing the evolutionary history of the species as a whole using the coalescent. I am also interested in linkage and QTL analysis, and have worked on problems involving microarray data as well. I have ongoing collaborations with Dr. Julie Wilder of Lovelace Respiratory Research Institute to study the murine immune response using QTL and microarray data and with Dr. Bill Beavis of the National Center for Genome Resources in the area of microarray data analysis. I also collaborate with Dr. Anne Stone, an anthropologist at Arizona State University, on phylogenetic studies of primate evolution and linkage studies of the genetic basis for tuberculosis susceptibility in Native South American populations.

Lee

My research interests include nonparametric function estimation, reproducing kernel Hilbert space methods in particular, classification, and statistical learning. In general, I am interested in developing methods and computational tools to cope with problems which feature a vast amount of data and complex underlying mechanism. To enhance our statistical understanding of machine learning methods and improve their interpretability and effectiveness for large scale high dimensional data, my current research focuses on feature selection, model selection, and their ramifications.

Lin

My research interests are in statistical genetics and genetic epidemiology. I currently focus on the development of statistical and computational methods for linkage analysis, association mapping, and the analysis of microarray gene expression data. The sort of data that render conventional methods infeasible, such as data from large families with complex relationships, is of a long-standing interest of me. Other important issues in mapping complex traits, such as genetic heterogeneity and multiple testing, are also of particular interest. Some of the methods that we are developing are based on Monte Carlo simulations utilizing Markov chain Monte Carlo methods and the method of sequential imputation. I am also collaborating with researchers in the Medical School in two genetic epidemiological studies. The first is a large scale genetic as well as longitudinal study to uncover genetic and environmental risk factors for systemic lupus erythematosus. The second study is on multiple sclerosis, in which we test candidate genes for their association with the disease. My collaboration with a group of researchers at the Florida State University to study the genetics of poodle epilepsy continues. We are currently investigating the mode of inheritance in standard poodles using the data that we have collected over the last seven years.

Liu

My research is primarily focused on statistical analysis of spatial data and informed by disciplines of spatial statistics, machine learning, remote sensing, and geographic information systems. Broadly speaking, I am interested in the development of efficient spatial analysis methods for the assessment and monitoring of natural resources and environmental changes. Specifically, my research has an emphasis on spatio-temporal data analysis and includes the following three areas: 1) Geocomputational algorithms for multi-temporal and multi-source data, 2) Bayesian spatio-temporal classification for land cover and land use change, and 3) Spatial pattern analysis and risk modeling for disease dynamics.

MacEachern

My current research focuses on several areas. The first is the development and application of nonparametric/semiparametric Bayesian models that incorporate covariate information in a natural sense. These probability models for a random distribution create a framework whereby a collection of distributions is allowed to evolve in a continuous fashion as a covariate changes. The second is the creation of Monte Carlo techniques for fitting models and exploitation of the output. The third is work on ranked set sampling, with an eye toward creating more robust inferential methods and eventually connecting ranked set sampling to Bayesian methods. This research is driven by a collection of problems arising across a wide range of application. Of special interest is application of psychological models to choice problems arising in the context of marketing.

Nagaraja

Areas of current research include: (i) order statistics, concomitants of order statistics, and record values, (ii) modeling, analysis and interpretation of heart period data, (iii) measuring agreement among instruments, (iv) correlation studies in repeated measure designs, (v) modeling duration of sleep, and (vi) several collaborative studies with medical researchers.

Notz

I am currently working in several areas of experimental design. These include problems in computer experiments, designs for correlated responses, and designs for crossover experiments. More specifically, I am focusing on the following.

1. Computer experiments. I am currently working with several students on problems in calibration of computer experiments, estimating quantiles in computer experiments, sequential design strategies for the global fit of GASP models to computer experiments, diagnostic tools for computer experiments, and incorporating qualitative variables into Gaussian process models for computer experiments.

2. Optimal experimental designs for correlated responses. This is joint work with Angela Dean and H-C Tsai.

3. Optimal experimental designs for crossover experiments. This is joint work with Angela Dean, Mausumi Bose, and Dongkwon Park. We are currently completing a paper that we plan to submit to Journal of Statistical Planning and Inference shortly.

Finally, I am revising the book Statistics Concepts and Controversies.

Özturk

My research involves drawing parametric, nonparametric and robust inference based on nonstandard sampling procedures such as ranked set sampling and its modifications. In parametric settings, my current research develops goodness-of-fit testing procedures for some common distributions. In nonparametric settings, I continue to develop distribution free-inference that is not affected by ranking error. In robust inference, I develop robust inferential procedure against imperfect ranking in ranked set sampling. I am also working on a major project, through collaboration with Professor Steve MacEachern, to develop estimation and testing procedures for the design of experiments for settings where the cost of each experimental unit is limited or expensive. Finally, Steve MacEachern, Doug Wolfe and I are writing a book Statistical Inference with Ranked Set Samples.

Pearl

One direction of my current research is in addressing statistical issues caused by the dependency structure of phylogenetic trees that seek to describe the evolutionary history of a group of organisms. Examples include the development of methods to explore the randomness in our knowledge of a phylogenetic tree; stochastic search methods to find good estimates of the phylogenetic tree; and new metrics in tree space to allow for a concise description of results. I am particularly interested in the applications of these techniques to the evolution of viruses and other disease causing organisms. Secondly, I have been active in many collaborative projects with biomedical scientists, including recent studies of the diagnoses and prognoses of brain tumors, and of the role of tobacco and alcohol in the development of oral cancers. A third area of research is in the development of new teaching strategies that provide an individualized match to student learning styles while reducing costs through the appropriate use of technology.

Peruggia

My research focuses on a combination of methodological and applied aspects. On the methodological front, my principal interests are in the areas of Bayesian modeling and Markov chain Monte Carlo estimation. I am especially interested in the development of numerical and graphical Bayesian techniques for exploratory data analysis, model building and selection, and in the related issue of constructing Bayesian diagnostics for assessing model fit and detecting the presence of outliers. With regard to estimation, I am interested in developing techniques for reducing the variability of estimators based on Markov chain Monte Carlo output, in the use of importance sampling techniques to accomplish modeling and diagnostic tasks, and in the development of Markov chain Monte Carlo algorithms for mixture models. On the applied front, I participate in several collaborative projects with colleagues and students at OSU and other research universities. Examples of ongoing collaborations include the development of Bayesian models and methods for the analysis of response time data, the modeling of state dependence in consumer purchase histories using epidemic-type point process models, and the analysis of sleep study data in patients with sleep apnea.

Santner

My research interests are in the design and analysis of experiments. Most recently, with Professors William Notz, Angela Dean, and OSU graduate students, and I am working to develop methodology for the design and analysis of "computer" experiments. Computer experiments are computer codes developed from mathematical models of input-output relationships. Historically, input-output relationships in agriculture, industry, and medicine have been studied using physical experiments. In some applications, such as determining the mechanical performance of prosthetic devices, input-performance relationships can also be studied using finite-element based computer codes. In prosthesis design, the inputs to codes are prosthesis geometry, the mechanical conditions of the patient's bone, and deviations from "ideal" surgical insertion parameters. In other applications, the inputs to codes represent operating conditions or engineering design specifications, field conditions, and unknown model parameters. We develop statistical methodology for screening, optimization, uncertainty analysis, and improved prediction of computer output that accounts for stochastic field conditions and deviations from idealized operating conditions or engineering design.

Shi

My research interests are statistical properties of machine learning algorithms, statistical methodology and computation on massive data sets, and applications of statistics in environmental science and geo- science. I have conducted research in building efficient and accurate statistical algorithms to detect clouds over polar regions using satellite data provided by NASA's Earth Observing System. In statistical methodology research, we studied the statistical properties of machine learning algorithms (Support Vector Machines and spectral learning algorithms) and applied them to the cloud detection problem. My long-term goal is to further study statistical methodology needed for large-scale dataset analysis and apply such methods in the context of statistical problems arising in geo- and environmental sciences.

Stasny

Most of my research is motivated by applications in the social sciences. In particular, I enjoy working with missing data in large-scale sample surveys and collaborating with faculty members from across the University in fields as diverse as Nursing, Human Nutrition, Geology, and Sociology. Current research on "Pathways to Overweight in Children", with Prof. Pam Salsberry (Nursing) and Prof. Pat Reagan (Economics) along with Statistics graduate students Shari Modur and Yi Liu, involves longitudinal models for both discrete and continuous data, accounting for missing data, to study what might be early indicators of obesity in children. With Prof. Mike Maltz (Sociology/Criminology) and Statistics graduate student Clint Roberts, I am working on "Imputing Monthly UCR Crime Data". This research involves developing methods to impute for missingness in the monthly FBI Uniform Crime Reports (UCR) from the 18,000 police reporting units across the US for the past 30 years. I am also working in Ranked Set Sampling with Prof. Doug Wolfe and our graduate students, Jessica Kohlschmidt, Chris Sroka, and Nader Gemayel. We are exploring problems on missing data in RSS, allocation in stratified RSS, and RSS for unequal set sizes.

Verducci

My focus is currently on statistical learning and data mining methods, with applications to drug discovery and biological networks. I’m directing dissertations for four graduate students: Li Yu, on detecting monotone association in subpopulations; Yushi Liu, on relating transport gene cluster profiles to the pituitary-hypothalamus-gonad system of salmonids; Shengjun Liu on data mining methods for finance; and Kimberly Walters, who will soon finish her dissertation on longitudinal designs with delayed treatment for the control group.

Wolfe

The area of ranked set sampling continues to dominate my research effort through joint work with Elizabeth Stasny and graduate students Nader Gemayel, Jessica Kohlschmidt, and Chris Sroka. We are currently working on the following problems in ranked set sampling settings: (1) missing data (Kohlschmidt), (2) optimal unbalanced RSS allocation in stratified survey sampling designs (Sroka), (3) randomly unequal set sizes (Gemayel), and (4) optimally sequential and partially sequential estimation of a cdf (Gemayel). We are also exploring ways to foster the application of RSS procedures in the collection and analysis of experimental data.

Xu

My current research has two main components: a) I am developing a set of theory and methodology of high dimensional density estimation in some general setups, such as extending the established density forecasting results to the unknown variance case and establishing asympotic results of predicitive densities in nonparametric regression. B) Although trained as a theoretical statistician, I am also trying to start some collaboration work with people in other fields. I am involved in the Initiative in Population Research at Ohio State. I am also working with Dr. Guofu Zhou in the Department of Finance at Washington University, trying to apply Bayesian decision theory to asset allocation problems.