1937

Автор(ы): 

Автор(ов): 

1

Параметры публикации

Тип публикации: 

Доклад

Название: 

Estimation of Marginal Density by Dependent Data

Наименование конференции: 

  • Stochastic Performance Models for Resource Allocation in Communication Systems

Город: 

  • Amsterdam

Издательство: 

  • CWI

Год издания: 

2006

Страницы: 

77-80
Аннотация
Huge data sets from the teletraffic industryexhibit many non-standard characteristics such as heavy tails andlong range dependence. The paper is devoted to the nonparametrickernel estimation of the univariate probability density functionby dependent time series data. The knowledge of univariate marginsis especially important for the bivariate analysis of data, e.g.,TCP-flow sizes and durations. Such analysis allows to evaluate thedistribution of the maximal throughput. It can be used forintrusion detection, too. It is known that the bias of a kernelestimate is the same for independent and dependent data. However,the variance is larger for the dependent case and depends on thecorrelation structure of the data. It is the idea to select such abandwidth of the kernel estimate to reduce the $MSE$ in the caseof dependent data. It is shown that the $MSE$ can converge forshort- or long-range dependent data at the close rate as the datawould under the assumption of independence.

Библиографическая ссылка: 

Маркович Н.М. Estimation of Marginal Density by Dependent Data / . Amsterdam: CWI, 2006. С. 77-80.