The asymptotic properties of the modified estimators of Gini's mean differences
The paper proposes a modified estimator Gini's mean difference, which is part of a class U-statistics based on the trimmed samples. It is shown that this estimate is asymptotically normal, has a limited influence function, it has a high efficiency for a normal distribution of observations. We assume that X1,...,Xn a random sample from a distribution function F(x) and we assume that has a density f (x) , x e R, X(),...,X(n) - ordered statistics of the original sampleX^...,Xn . Let T(F) , F e 5 common functional that characterizes the scale parameter, which describes the degree of dispersion of the study of a random variable X . We consider the functional which is defined as 1 F (1-a) Aa(F)= 2 J Jx - y\dF (x)dF (y), 0 a -estimators with other estimates of the scale parameter for Gaussian model with s -fixed proportion of contamination, and proposed an adaptive version A a -estimators for which the parameter a characterizing the proportion of "trimmed" of the original sample, selected on the basis of information contained in the original sample using a sample estimate functional, characterizing the degree of "heavy tails" of the distribution function of the random variable X under study.
Keywords
масштабный параметр, робастные оценки, функция влияния, средняя разность Джини, U-статистики, адаптивные оценки, scale parameter, robust estimation, influence function, asymptotic relative efficiency, adaptive estimatorsAuthors
Name | Organization | |
Shulenin Valery P. | Tomsk State University | shvp@fpmk.tsu.ru |
References

The asymptotic properties of the modified estimators of Gini's mean differences | Vestnik Tomskogo gosudarstvennogo universiteta. Upravlenie, vychislitelnaja tehnika i informatika – Tomsk State University Journal of Control and Computer Science. 2016. № 4(37). DOI: 10.17223/19988605/37/8