Computational Optimization Methods in Statistics, Econometrics and Finance

Partners involved
Univ. of Giessen (Local Organizer), Imperial College, Univ. de Genève.

Date and location:
15/10/2008 - 19/10/2008 in Giessen, Germany.

Motivation and Objectives

A more general application of optimization heuristics in statistics, econometrics and finance requires a different interpretation of the results. In particular, researchers have to communicate in a comprehensive but informative way the results of several thousands of examples. One way to accomplish this goal is by means of theoretical convergence results for the estimators and the numerical procedures if feasible. Another way is to use statistical tools such as response surface or regression analysis to provide empirical information on the results. The tutorial will introduce concepts and methods for analysing the results of computationally intensive methods, in particular using optimization heuristics. Furthermore, it will teach standards for presenting the results in a way accessible for other researchers.

Organization

Two days with six lectures of 90 minutes and one afternoon “hands on” session.

1. Classical convergence analysis for econometric estimates and Monte Carlo simulation.
2. Convergence results for optimization heuristics.
3. Combining convergence of estimators and optimization methods – an application to GARCH-models.
4. Extreme value theory as a method for analysing empirical results.
5. Extreme value analysis applied to the results of repeated applications of optimization heuristics.
6. Presentation of empirical results by means of response surface analysis and regression analysis including links to the convergence properties.
7. “Hands on” session in the computer laboratory.