Computer simulation models are widely used in diverse scientific disciplines for understanding the relationships between variables; examples include models of pharmacokinetics, ecosystems, epidemiology, climate, and biological populations. Evidence related to such models and their predictions arises both from observed data and from the opinions of experts whose expertise is amassed over years of experience and study rather than being based on any particular set of data. Statistical methods to combine these two types of information will be developed to help researchers make overall inferences about simulations, particularly future projections and the uncertainties therein. The statistical innovations developed will also deal with the difficult situation when expert opinions conflict. These methods will be flexible, powerful new tools for non-statisticians and scientists in diverse fields to enhance what they learn from their own simulation modeling.
Several instructional initiatives are also planned, with the primary goal of student recruitment and interdisciplinary cross-fertilization. They are designed to nurture an early enthusiasm for applied statistics and then to serve as a conduit to direct students towards undergraduate and graduate degrees in the field, as well as to provide a pipeline of statistically literate students of the scientific workforce. The proposed research will also enrich graduate education by providing opportunities for students in several disciplines to interact and participate in relevant research. A key area of student participation will be the development of an interactive computer program that can be used by scientists from diverse disciplines engaged in simulation modeling to draw statistically sound inferences from their simulation experiments. In this sense the project should benefit the broader scientific community.
