DIMACS Computational and Mathematical Epidemiology Seminar Series

Title: SVM in Analysis of Cross-Sectional Data

Speaker: Dmitriy Fradkin, DIMACS

Date: April 4, 2005 2:30 - 3:45 pm

Location: DIMACS Center, CoRE Bldg, Room 431, Rutgers University, Busch Campus, Piscataway, NJ


In cross-sectional studies, one of the basic designs in observational epidemiology, measures of disease occurrence and variables of interest are obtained, at a single point in time, from the whole population or a representative sample. The collected data is then analyzed to detect "risk factors". This focus on identification of important variables distinguishes the task of analyzing cross-sectional data from other applications of classification methods, where the goal is to build a good predictive model.

This talk will discuss the potential of SVM to provide an alternative to logistic regression (currently the method of choice in epidemiology), and the importance of statistical model validation.