Sensitivity Analysis of Simulation Experiments: Regression Analysis and Statistical Design

Abstract

This tutorial gives a survey of strategic issues in the statistical design and analysis of experiments with deterministic and random simulation models. These issues concern validation, what-if analysis, optimization, and so on. The analysis uses regression models and least-squares algorithms. The design uses classical experimental designs such as 2k−p factorials, which are more efficient than one at a time designs are. Moreover, classical designs make it possible to estimate interactions among inputs to the simulation. Simulation models may be optimized through response surface methodology, which combines steepest ascent with regression analysis and experimental design. If there are very many inputs, then special techniques such as group screening and sequential bifurcation are useful. Several applications are discussed. (This abstract was borrowed from another version of this item.) (This abstract was borrowed from another version of this item.) (This abstract was borrowed from another version of this item.) (This abstract was borrowed from another version of this item.)