Held on October 20, 2005 at the Renaissance Chicago North Shore Hotel.
The Program consisted of three presentations:
- Sequential Logistic Regression: Modeling Risk Factors and Child Outcomes Anita Ross, Ph.D., North Park University
- Adaptive Tests of Significance and Confidence Intervals Tom O’Gorman, Ph.D., Northern Illinois University
- An Algorithm for Local Modeling of Protein Complexes Denise Scholtens, Ph.D., Northwestern Medical School Department of Preventive Medicine
__________________________________________________________________________________________________________________________________________________________
Sequential Logistic Regression: Modeling Risk Factors and Child Outcomes by Anita Ross, Ph.D., North Park University
Biographical Background Anita A. Ross, Ph.D., received her Ph.D. from De Paul University in 1990. She has worked as a statistician and research analyst in the general area of health care treatment outcomes for many years. At the same time she taught statistics and research courses at several area colleges and universities. She is currently an associate professor in the School of Adult Learning at North Park University, Chicago Illinois. She is responsible for the mathematics, statistics and computer information systems courses offered through the School of Adult Learning. She continues work as an independent statistical consultant. She has been a long-term member of the NIC, and has served on the executive committee as newsletter editor since 2000.
Abstract This presentation will overview logistic regression analyses and present the results of a study in the area of social-science research. We will begin by going over basic mathematical concepts underlying logistic regression, presenting an example of the simplest case of logistic regression with a single (binary) predictor. The process of inference and the interpretation of the odds-ratio will be explained. A multivariate example will be described in which logistic regression was used to explore the validity of a model of the association between certain risk factors and short- and longer-term outcomes for high-risk children.
The goal of these analyses was to explore risk factors associated with short- and longer-term child outcomes. The theoretical model guiding the analyses posited that maternal history of trauma, in the form of either loss of one’s parent before age 18, or a history of abuse or neglect in one’s childhood would be associated with compromised maternal emotional status, evidenced by misuse of alcohol or drugs, clinical symptoms of depression, above average report of psychosomatic symptoms, or hospitalization for psychiatric or substance abuse problems. Maternal compromised emotional status would be associated with dissatisfaction and/or problems in the environment of the family and/or neighborhood. Dissatisfaction and/or problems in the family/neighborhood environment would be associated with increased risk for domestic violence. Domestic violence would be associated with poor short-term outcomes for the child. Poor short-term outcomes for children include either having a founded report of child abuse or neglect in the household or experiencing two or more of the following: presence of alcohol or drugs in the body at birth, cognitive delay, a health problem or a hospitalization. Poor short-term outcomes would be associated with poor longer-term outcomes including health problems, cognitive and behavioral developmental delay.
Sequential logistic regression analyses were used to explore the validity of this model of relationships among risk factors and poor short- and longer-term child outcomes. General support for the model was found, but diagnostic examination of the results of the logistic regression analyses revealed some weaknesses.
__________________________________________________________________________________________________________________________________________________________
Adaptive Tests of Significance and Confidence Intervals by Tom O’Gorman, Ph.D., Northern Illinois University
Biographical Background Tom O'Gorman is an Associate Professor of Statistics at Northern Illinois University. He received his Ph.D. in Biostatistics from the University of Iowa where he worked as a statistical consultant. He also worked as a statistical consultant for the Southwestern Bell Telephone Company. In the last few years his research has been focused on methods to improve the performance of basic statistical procedures.
Abstract Accurate systems biology modeling requires a complete catalog of protein complexes and their constituent proteins. We discuss a graph-theoretic/statistical algorithm for local modeling of protein complexes using data from affinity purification-mass spectrometry experiments. The algorithm readily accommodates multicomplex membership by individual proteins and dynamic complex composition, two biological realities not accounted for in existing topological descriptions of the overall protein network. A penalized likelihood approach guides the protein complex modeling algorithm. With an accurate complex membership catalog in place, systems biology can proceed with greater precision.
__________________________________________________________________________________________________________________________________________________________
An Algorithm for Local Modeling of Protein Complexes by Denise Scholtens, Ph.D., Northwestern Medical School, Department of Preventive Medicine
Biographical Background Dr. Scholtens received a bachelor's degree in mathematics with from Wheaton College in Wheaton, IL in 1997 and a PhD in biostatistics from Harvard University in 2004. In 2004, she joined the Department of Preventive Medicine faculty at Northwestern University and is a biostatistician for the Robert H. Lurie Comprehensive Cancer Center. Dr. Scholtens is interested in the development of methodology for the analysis of high-dimensional bioinformatics data and is an active contributor to the Bioconductor project (www.bioconductor.org). She is currently working on local modeling of global protein interactome networks, the joint analysis of these networks with gene expression data, and global measures of network topologies that incorporate experimental design and false-positive/negative observations of edges.
Abstract Accurate systems biology modeling requires a complete catalog of protein complexes and their constituent proteins. We discuss a graph-theoretic/statistical algorithm for local modeling of protein complexes using data from affinity purification-mass spectrometry experiments. The algorithm readily accommodates multicomplex membership by individual proteins and dynamic complex composition, two biological realities not accounted for in existing topological descriptions of the overall protein network. A penalized likelihood approach guides the protein complex modeling algorithm. With an accurate complex membership catalog in place, systems biology can proceed with greater precision.