Melanie Wall, PhD
Professor of Biostatistics (in Psychiatry)
Dr. Wall is a biostatistician with a sustained record of focused, innovative data analysis methods research in the areas of latent variable modeling, spatial, and longitudinal data analysis. Before joining the Psychiatry Institute at Columbia University in July 2010, she was on faculty for 12 years at the University of Minnesota School of Public Health. She has made distinctive developments in latent variable modeling including a series of methods which allow for nonlinear relationships to be examined among latent variables and an innovative line of research that incorporates latent variables into spatial (i.e. geographical) data analyses. She has a sustained record of securing funding as principal investigator for her methodological research: “Minnesota Center for Excellence in Health Statistics” (NCHS 1999-2004), “Latent variable models and methods for behavioral health data in public health” (R01 NIH-NIAAA 2005-2009), and recently “Multilevel latent class and social network models for observational adolescent obesity data” (U01 NIH-NCHD 2009-2011).
Dr. Wall’s unique expertise in the areas of factor analysis, latent class models, structural equation modeling as well as spatial data modeling have lead to many fruitful collaborations with health researchers. For example, she developed a new methodology for examining associations between continuous and categorical latent variables which was used for linking body image with eating disorder risk, and proposed an innovative solution using hidden Markov models to a problem of testing the efficacy of an alcohol treatment programs using medical claims records. Dr. Wall was integral in developing a new way of clustering parenting behaviors around child obesity, classifying state policies regarding school nutrition and physical activity, describing systematic obstacles to good care among nurses and carefully measuring pain in the context of TMJ and back problems.
My research interests include: factor analysis, latent class models, item response theory, structural equation modeling as well as spatial and longitudinal data modeling.
I currently have a U01 from NIH-NCHD entitled "Multilevel latent class and social network models for observational adolescent obesity data" to develop and examine new statistical models and methods incorporating latent variables, social networks and causal modeling that simultaneously incorporate multiple variables from multilevels of influence in order to inform a richer understanding of the etiology of on adolescent obesity.
Social network methodology is a new area of research for me but builds directly on my work in spatial data modeling and represents a cutting edge new direction for studying the group dynamics of health behaviors and outcomes.