Biostatistics
Christine M. Waternaux, Ph.D., Director
Steven P. Ellis, Ph.D., Research
Scientist V
Xinhua Liu, Ph.D., Biostatistician
Todd Ogden, Ph.D., Biostatistician
Eva Petkova, Ph.D., Senior
Biostatistician
The Biostatistics department provides investigators of the New York Psychiatric Institute (NYSPI) access to and training in state-of-the-art statistical techniques, as needed for optimal use of their research data. This includes: developing and applying new methodology for the design and analysis of psychiatric studies, participating in major funded research projects, teaching statistics to researchers and fellows of training grants, and providing consultations on data-analytic and design issues for grant submission and manuscript preparation.
Department Staff
The Biostatistics department has four senior faculty members: Dr. Christine Waternaux, Director, and Drs. Eva Petkova, Steven P. Ellis, and Xinhua Liu. Drs. Waternaux, Petkova and Liu also hold academic appointments in the Division of Biostatistics of the Columbia University School of Public Health (CUSPH), and Dr. Ellis has a faculty appointment at the Columbia University Department of Psychiatry. In July 2000, Todd Ogden, Ph.D. was hired to work full-time as a Biostatistician for the Core Grant to Enhance Neuroscience Transfer (GoGENT). Dr. Ogden's work focuses on biostatistical issues related to brain imaging and interpretation. Guoguang (Julie) Ma, a doctoral student in the Ph.D. program of the Division of Biostatistics at the Columbia University School of Public Health, assists the department staff as a Research Associate. Her duties include complex statistical modeling and non-standard statistical analysis, consulting under Biostatistics staff supervision, manuscript preparation, training of data analysts and network management. Jianfeng Chen, also a Ph.D. candidate of the Division of Biostatistics at the Columbia University School of Public Health, was hired in August to work on grants awarded to the department. John Scott Lucas provides clerical support for the division and coordinates the weekly NYSPI Biostatistics Workshop.
Methodological and Applied Research Projects
Methodological research in the department focuses on problems of particular relevance to psychiatric studies. Areas of expertise include analysis of longitudinal studies with generalized linear models and variance components models, brain imaging, survival analysis, models for the analysis of categorical data, methods for dealing with missing data including data with non-ignorable non-response mechanism, causal models, empirical Bayesian techniques, methodology for combining information from different sources (e.g., meta-analysis), classification methods with recursive partitioning tree models, resampling techniques (e.g., bootstrapping), wavelets applications, and other computationally intensive methods.
Modeling of Functional Cognitive Decline with Prevalent Cohort Data Drs. Waternaux and Liu continued to collaborate with Jeanne Teresi, Ed.D., Ph.D. to extend previous work on the modeling of functional cognitive decline with prevalent cohort data. The method is applied to compare the cognitive decline in the elderly residing in special care units with those residing in non-special care units in the New York site of the National Aging Study (article published in Statistics in Medicine, vol. 19).
Empirical Bayes Methodology for Analysis of Drug Safety and Interactions Dr. Waternaux is a special Consultant to the Food and Drug Administration (FDA) and participates in research on the safety of drugs and combination of drugs. She is implementing a "data mining" technique to analyze data from research on psychotropic drugs and their interactions by gender using the FDA post-marketing database heretofore unanalyzed. She is also applying Bayes and Empirical Bayes methods for performing meta-analyses on this database.
New Methodologies for Estimation of Speed of Action of Antidepressants and Comparison of Therapeutic Strategies In a continuing collaboration with the researchers from the Depression Evaluation Service (DES), Dr. Petkova and Julie Ma have been developing a methodology for assessing the onset speed of antidepressants. A closely related problem is the one of devising strategies for switching non- responding patients from one antidepressant to another. This problem involves characterization of these strategies with respect to their expected losses. Dr. Ogden is working on extending methods of multivariate classification to functional data analysis setting. Eventually these methods may be applied to classifying subjects based on their response levels to antidepressants or placebos.
Epidemiological Study of Eating Disorders in a Non-referred Adolescents Population The Department of Health & Human Services has awarded a research grant to Dr. Petkova to investigate the potential risk and protective factors for eating disorders. One of the goals of the study is to identify factors that may be useful in screening for typical and atypical eating disorders in pediatric practice. Another goal is to begin to understand the etiology of these disorders and to generate hypotheses for future investigation. In collaboration with the Department of Child Psychiatry and under the supervision of Eva Petkova, Jianfeng Chen has undertaken the analysis of sample of over 5000 non-referred adolescents. Statistical methods appropriate for the questions addressed include statistical modeling using CART, graphical methods for identification of conditional independence relationships, and causal modeling.
Brain Imaging Dr. Ellis continued work as Principal Investigator for an NIMH-funded project, "Statistical Analysis of Human Neurochemical Maps." This project focuses on developing statistical methods for the analysis of autoradiographic brain images. Dr. Ellis performs this work in close collaboration with Victoria Arango, M.D. and J. John Mann, M.D.
As part of his work with the CoGENT, Dr. Ogden has been working to improve and automate some aspects of analysis of Positron Emission Tomography (PET) imaging, using non-linear optimization in maximum likelihood estimation (Principal Investigator, J. John Mann, M.D.). Extending this research involves developing new diagnostics for goodness of fit for determining optimal subsets of data for efficient estimation.
Dr. Ogden is an authority on wavelet applications in statistics, and he is teaching a class on the subject at Columbia University School of Public Health. He is currently working on modeling random functional effects with multiple replications with wavelet thresholding. This research may yield practical applications in the field of brain imaging.
Collaborative and Consultative Activities
Department staff have been very active in their collaborative and consultative efforts, especially with the Departments of Neuroscience, Therapeutics, Biological Psychiatry, Clinical Psychopharmacology, Clinical Psychobiology, Geriatrics, Gerontology, and Substance Abuse. In 2000, department staff provided input on design and data-analytic strategy for over 20 grant submissions.
Consultations with Center Grants and Program Project Grants As Senior Statistician for the Huntington's Disease Research Group (Principal Investigator, Nancy Wexler, Ph.D.), Dr. Waternaux has been analyzing the large longitudinal data set collected by Dr. Wexler in Venezuela over the last sixteen years. Dr. Waternaux is also the Director of the Statistics and Computing Core of the MHCRC for the Study of Suicidal Behavior (Director, J. John Mann, M.D.). Dr. Ellis is Associate Director of the Statistics and Computing Core.
Dr. Petkova is Principal Investigator of the Biostatistics, Database Management, and Networking (BaDMaN) Core for the Core Grant to Enhance Neuroscience Transfer (CoGENT), (Principal Investigator, Jack Gorman, M.D.). This unique NIMH-funded project is designed to coordinate efforts, reduce redundant work, and facilitate creative interaction between ten related research projects. (Principal Investigators of Core Grants: J. John Mann, M.D.; Gerry Bruder, M.D.; Harold Sackeim, M.D.; Steven Roose, M.D.; Brian Fallon, M.D.; Michael Liebowitz, M.D. and Patrick McGrath, M.D.).
Dr. Ogden's work with the BaDMaN Core includes investigating a variety of ways to improve and evaluate motion correction (registration) of PET images. Utilizing data from the Department of Neuroscience (Principal Investigator, J. John Mann), he has developed diagnostics for severity of motion artifacts based on residuals from the model fits. He is also studying the practicality of wavelet denoising in motion correction.
Other Consultations Dr. Liu is analyzing data on DSM-IV alcohol dependence and sustained reduction in drinking from a community sample, (Principal Investigator, Deborah Hasin, M.D.) She also provided CART analysis for a study of recent and remote suicide attempters compared to non-attempters (Principal Investigator, J. John Mann, M.D.). Dr. Liu continued to analyze longitudinal data collected in Yugoslavia, (Principal Investigator, Joseph Graziano, Ph.D.). This analysis focuses on two areas: lead exposure and motor functioning in 4-1/2-year-old children, and the association between bone lead and blood lead measures. All of these studies have produced one or more papers, either published, in press, or in submission.
Education and Training
Statistics Workshop As part of its training activities, the department offers the NYSPI Biostatistics Workshop, a series of seminars covering statistical methodology, data management, and the use of advanced statistical software. Invited speakers illustrate statistical and data management issues based on examples from on-going research at the Institute, explore emerging methodologies, or present relevant papers recently published in the statistical literature. These weekly seminars are open to all Institute investigators and their staff members, as well as students from the Columbia School of Public Health.
Columbia-Penn Forum on Statistics in Psychiatry On March 28, 2000, the Department of Biostatistics attended the Second Annual Columbia-Penn Forum on Statistics in Psychiatry, which was hosted at the University of Pennsylvania's main campus in Philadelphia. Dr. Petkova presented a paper, entitled, "Mixed Effects Models for Estimation of the Total and Direct Effects of Residence in Special Dementia Care Units on Function Using Clustered Longitudinal Data." The department will host the Third Annual Columbia-Penn Forum on Statistics in Psychiatry on April 17, 2000, in New York City. Todd Ogden, representing NYSPI and Columbia University, will present the paper, "Some Current Statistical Issues in Brain Mapping".
BaDMaN Workshops As part of the Department of Biostatistics' work with the Biostatistics and Data Management and Networking (BaDMaN) Core, Dr. Petkova organized the BaDMaN Workshops to further the exchange of information, foster collaboration, and facilitate the development of new ideas between NYSPI researchers, biostatisticians, and data managers. This bi-monthly seminar began in the Fall semester of 2000. Topics covered included: "Statistical methods for looking at data, checking the assumptions of underlying t-test, ANOVA, MANOVA and linear regression, and model selection," by Eva Petkova; "Identifying placebo responders," by Todd Ogden; and "A demonstration of web-based research tools from a web-based project management system for the Jerusalem study," by Howard Andrews, Gerry Segal, and Chandresh Shah.
Department World Wide Web Home Page
The World Wide Web home page for the Biostatistics department at NYSPI provides an easily accessible resource for NYSPI investigators and other interested parties. The department web site can be found at http://hopi.cpmc.columbia.edu/nyspi/depts/biostat/
Participants in the 3rd
Annual Columbia-Penn Forum on Statistics in Psychiatry, Tuesday, April
17, 2001, held at the New York State Psychiatric Institute. Standing
L-R: Penn Guest; Warren Bilker, Ph.D., U. Penn; Eva Petkova, Ph.D.,
NYSPI/CU; Xinhua Liu, Ph.D., NYSPI/CU; Todd Ogden, Ph.D., NYSPI/CU;
Penn Guest; Dan Heitjen, Ph.D., NYSPI/CU; Steven Ellis, Ph.D., NYSPI/CU.
Seated, L-R: Penn Guest; Penn Guest; Penn Guest; Tom Tenhave, Ph.D.,
U.Penn.