An Empirical Comparison of Conventional Moment and L-Moment Based Power Method Polynomial Transformations

• In Event: Roundtable Session 41
  In Roundtable Session: Simulation Studies

Abstract

Conventional moment-based power method polynomials are often used for simulating non-normal distributions. However, conventional moment-based estimators can (a) be substantially biased, (b) have high variance, or (c) be influenced by outliers. Thus, a comparison between conventional and L-moment-based power method polynomials is proposed. Specifically, a system of equations is provided for determining coefficients for specified L-moment ratios for standard normal based polynomials. The comparison reveals that L-moment estimators are (a) nearly unbiased (for any sample size), (b) have relatively small variance, (c) insensitive to outliers, and thus (d) superior to conventional moment-based polynomials. It will be shown at the time of presentation how multivariate non-normal distributions can be simulated with specified L-moments and L-correlations

Authors

• Mohan Dev Pant, The University of Texas - Arlington

• Todd Christopher Headrick, Southern Illinois University - Carbondale