Comparison Between Robust Trimmed and Winsorized Mean: Based on Asymptotic Variance of the Influence Functions
DOI:
https://doi.org/10.53555/ms.v3i12.227Keywords:
Trimmed Mean, Winsorized Mean, Influence Function, Monte Carlo SimulationAbstract
robust trimmed mean and winsorized mean has been compared in terms of influence function under the situation when a small change occur in the underlying symmetric distribution. The behavior of the two robust estimators have been compared through the asymptotic variance of the influence functions of the corresponding estimators. A Monte Carlo simulation studies has also been conducted to examine how asymptotic variance the influence function of the two robust estimators behave with the variation of the amount of trimming as well as with various the sample sizes. The simulated result revealed that the asymptotic variance of the influence function for both robust estimators increases when the amount of trimming increases but having lower trend for the estimator winsorized mean. That is, the estimator winsorized mean provides more efficient as well as robust result compared to the estimator trimming mean.
Downloads
References
Hampel, F. (1974). The influence curve and its role in robust estimation, J. Am. Statist. Assoc. 69: 383393.
Huber, P.J. (1981). Robust Statistics. Wiley, New York.
Hampel et al. (1986). Robust Statistics: The Approach Based on Influence Functions. Wiley, New York.
Grubs, F.E. (1969).Procedures for detecting outlying observations in samples.Tecnometrics, 11, 1-21.
Hawkins, D., 1980. Identification of Outliers, Chapman Hall.
Johnson, R., 1992. Applied Multivariate Statistical Analysis, Prentice Hall.
