Working with the Environmental Epidemiology group, as well as taking course work within both departments helped me to identify my research topic, latent variable models for multiple outcomes. The idea is that there is an underlying latent outcome that cannot be measured which is directly related to the exposure of interest. Consider the study of birth defects where exposure causes a "syndrome". This is a setting where disease is characterized by the presence of a wide variety of different kinds of effects, any one of which may not be serious, but where the configuration together suggests a severe effect. Fetal alcohol syndrome, and Down's syndrome are two examples. This project is helping me to develop many skills including: goal setting, organization, and the ability to communicate on a technical level.
Joe Hogan writes: The idea for my thesis originated with my participation in a summer project after my first year of study. The AIDS Clinical Trials Group (ACTG) investigates treatment issues for people with HIV infection and with AIDS. The trial on which I worked studies the dose effects of AZT on neuropsychological development in pediatric AIDS patients. My primary responsibility that summer was to help Rich Gelber, Associate Professor of Biostatistics, and Nuala McGrath, Biostatistician with the SDAC, prepare an interim summary of the progress of the clinical trial. I performed basic statistical summaries, fit repeated measures models, and classified missing data patterns. Those analyses provided a useful summary of the data, but the models used to explain the treatment effects made very naive assumptions about the connection between the outcome values and the reasons some of those values are missing.
I became aware of some more sophisticated missing data methods in Nan Laird's Longitudinal and Multivariate Data Analysis class and through various discussions with her during the summer following her course. Later, she agreed to be my thesis advisor. The plan of my thesis is to develop and evaluate methods which account for dropout processes when making treatment comparisons and to apply the methods to the AIDS trial I was working on. Because Nan has spent the better part of her career studying missing data problems, I consider myself fortunate to be working with her.
Last modified $Date: 1994/12/01 20:23:33 $ by Evelyn Ophir, email@example.com