WELCOME TO RUSSELL BARBOUR STATISTICS
My specialty is the analysis of data that does not necessarily fit the assumptions of standard statistics because of correlation, or non-independence of the data due to space, time or other
factors. My research has developed novel applications of
Bayesian methods, Generalized Linear Models (GLM) and
associated Generalized Estimating Equations (GEE). I also do standard statistics such as ANOVA and regression analysis.
I have experience as an expert witness in medical risks and probabilities.
I have a special focus on Geographic Information Systems (GIS) , spatial regression and other spatial statistics.
I have worked extensively on USAID funded projects in design and statistically based monitoring and evaluation.
Dr. Barbour's research has focused on programming Splus, SPSS, STATA and R software for novel statistical applications in epidemiology and ecology. He has also done extensive research on geographic information systems (GIS) for both mapping and spatially explicit analysis and spatial statistics. His work has focused on data sets that present problems in terms of correlation, non-independence and other violations of the usual assumptions in most statistical analysis. He also works on methods of dealing with missing data through imputation and/or Bayesian modeling.
As Research Associate in Applied Mathematics at the Vector Ecology Laboratory of Yale School of Medicine Dr. Barbour applied spatial statistics and Artificial Neural Networks (ANN) to problems in vector borne disease here in the United States. Dr. Barbour used these evolving methods to integrate climate and microclimate data into probability assessments of human risk of Lyme disease and West Nile Virus.
In his current position as Co-Director for the Interdisciplinary Research Methods at the Center for Interdisciplinary Research on AIDS, (CIRA) at Yale School of Medicine, Dr. Barbour does research on application of statistical modeling, Bayesian analysis, and spatial statistics to HIV and hepatitis risk and transmission dynamics.
Most recently he has completed a spatial analysis of HIV risk factors among intravenous drug users in St. Petersburg, Russia. In October 2007 he taught a seminar course at CIRA on the application of Bayesian Methods to HIV research at Yale School of Medicine.?
In 2005, 2006, 2007 and 2008 Dr. Barbour taught spatial statistics, Bayesian methods and Generalized Estimating Equations at the Psychology Faculty of the State University of St. Petersburg and the affiliated Biomedical Center, where he holds the position of lecturer.
Dr. Barbour has various teaching and training responsibilities at Yale School of Medicine as well. These courses and seminars include geographic information systems (GIS), spatial statistics, Bayesian methods and training in R software.
As a member of the Board of Trustees of the non-profit organization Medical Care Development, Dr. Barbour continues to be involved in international health projects in South Africa, Equatorial Guinea, Sudan and Madagascar. He also continues his interest in wildlife conservation and has served as a member of the International Advisory Board of the Cape Peninsula National Park in South Africa and is a member of the Development Committee of the Beza Mahafaly Special Reserve in southwestern Madagascar.
Previously Dr. Barbour did design and evaluation of USAID funded projects in Africa, including Chad, Madagascar, Sudan, Somalia , Guinea Niger and other African countries. Clients included AFRICARE, USAID, Medical Care Development, Volunteers for Technical Assistance and other non-profit organizations.
EDUCATION:
Yale University, Ph.D
George Washington University, BA
RECENT POSITIONS:
2006 - Present
Co-Director, Data Management & Statistical Analysis, Center for Interdisciplinary Research on AIDS (CIRA), Yale University School of Medicine, New Haven, CT
2000
Co-Director, field training course in Landscape Epidemiology, Yale School of Forestry and Environmental Studies, New Haven, CT
1998 - 2004
Research Associate in Applied Mathematics, Vector Ecology Laboratory, Yale School of Medicine, New Haven, CT
1989 - 1995
Fellow / Lecturer, Yale School of Forestry and Environmental Studies, New Haven, CT
1992
Research Fellow, Institute of Forestry, Pokara, Nepal
1980 - 1989
Consultant in African Ecology and Agriculture
While standard statistical methods such as ANOVA and linear regression are appropriate for data that are normally distributed and fit other basic assumptions such as homoscedasticity of standard errors, in natural systems such as Epidemiology and Ecology one often finds correlation, non-independence, and heterscedasticity of standard errors that may mask statistical inference. The violation of these model assumptions are often treated as a nuisance that are conjured away through various statistical expedients such as removing data until independence is achieved.
In contrast Dr. Barbour has developed and applied statistical methods that use correlation, clustering and adjustment of standard errors to draw further inference from such data sets. The application of spatial statistics, generalized estimating equations (GEE), Social Network Analysis and Bayesian modeling can not only overcome such problems, but provide further insights into the systems under study and underlying relationships.
Dr. Barbour also works extensively with missing data. Many years of research are often frustrated by lacuna in key variables. Standard imputation methods such as averaging, "hot deck" and regression can introduce bias. There is growing agreement among statisticians that Bayesian modeling is the most robust statistical method for dealing with missing data. Using the winBugs program Dr. Barbour applies Gibbs sampling and Markov-Bayes simulations to over come incomplete data sets and produce statistically robust results.
