New Nonparametric Methods in Risk Analysis Based on Resampling Techniques and Empirical Simulation
Leon Emry Borgman · 2005
There has been an explosion of development in new nonparametric methodology in the last several decades, as workers in statistics sought to escape the tyranny of parametric and normality assumptions for problems that were clearly often not Gaussian and not easily described by the usual standard distribution functions. Methods have been developed which allow one to proceed from raw data sets directly to simulations of possible future new data with a bare minimum of parametric assumptions. Many of these techniques are now being introduced into engineering practice and decision-making. The presentation reviews the univariate and multivariate methods growing out of the resampling procedures, and their extensions for use in extremal statistics and the estimation of risk.