************************************************************************* Department of Mathematical Sciences The Johns Hopkins University SEMINAR ************************************************************************* Langche Zeng April 04, 2002 Department of Political Science 304 Whitehall Hall George Washington University Refreshments: 3:30 PM Seminar: 4:00 PM ************************************************************************* QUANTITATIVE POLITICAL METHODOLOGY ************************************************************************* ABSTRACT Like methodology subfields in the other social sciences, political methodology serves as a bridge between the subject field and the mathematical sciences, and aims at providing political scientists with useful mathematical/statistical tools for improving empirical political analysis. Political methodology is a small subfield relative to, for example, econometrics, but political methodologists tend to come with diverse background training, bringing interesting interdisciplinary perspectives to their work. As such, political methodologists acquire new tools both from importing and adapting existing techniques from a variety of other fields, and by creating their own methods, motivated by special features of political data, that often prove useful outside of political science, such as in economics, sociology and medicine/public health. This talk gives some recent examples of these endeavors, such as the use of machine learning methods for improving the forecast of international conflict and state failure, the development of methods for analysis of rare event data and the study of sources of bias in estimating causal effects with non-experimental data. The talk also discusses perspectives of improving prediction and causal inference, the central concerns of empirical political science, with graphical methods and models. The talk is based on joint work with Gary King, Harvard University.