Seminar in Financial Mathematics & Probability

Georg Ch. Pflug Department of Statistics and Decision Support Systems (ISDS) University of Vienna, Austria Stochastic optimization: From statistical data to optimal decisions Monday, April 30, 2001 3:15pm Pierce 116

Abstract:   Uncertainty is present in nearly all cases of optimal decision making: Financial planning, transportation and communication, network design, etc. Decision making under uncertainty uses two methodologies: The methodology of statistics to deal with probabilistic modeling and information extraction from data, as well as the methodology of optimization for finding the best out of several alternatives. Stochastic optimization builds the bridge between statistics and optimization: We show how large sample (asymptotic) approximations work in statistical estimation as well as in stochastic optimization to get statements about the quality of the decisions (confidence sets both for optimal values and for optimizers). As illustration, we present problems of financial management, in particular one- or two-stage portfolio management problems. Some statistical location parameters, as median, quantiles, etc. as well as dispersion parameters like the semivariance or mean absolute deviation appear in a natural way as objectives or constraints in portfolio management. This establishes a further link between statistics and stochastic optimization.

Coffee and refreshments will be available starting at 3:00pm. For additional information contact Khaldoun Khashanah  or Darinka Dentcheva.