Use of Color in the Output Analysis of Statistical Simulations, and Analysis of Estimators of Serial Correlation

摘要

The use of color in the organization and analysis of the output of multifactor statistical simulations is investigated with the computer package SIMTBED (A Simulation Test Bed). Updating of this system to the current technology of color line printers is performed. It is shown how color can be used to code some factors in a multifactor simulation, compacting the output and enhancing analysis. An application to the analysis of the lag one serial correlation of normal and non-normal time series using four estimators (moment, maximum likelihood, robust regression, and the Cressie estimator) is provided as a demonstration of the uses of SIMTBED in statistical simulations. These estimators are examined for robustness and asymptotic bias, as well as relative behavior with various sample sizes. It is shown that for some time series the robust estimators of serial correlation are not acceptable due to bias and other considerations. (KR)