TY  - CHAP
AV  - none
ID  - eprints1580
TI  - Analysis of a Hurst parameter estimator based on the modified Allan variance 
UR  - http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6503362&isnumber=6503052
KW  - Hurst parameter
KW  -  long-range dependence
KW  -  self-similarity
KW  -  modified Allan variance
KW  -  parameter estimation
KW  -  wavelets
KW  -  fractional Brownian motion
EP  - 1721
PB  - IEEE
T3  - GLOBECOM (ISSN 1930-529X)
SP  - 1716
T2  - 2012 IEEE Global Communications Conference (GLOBECOM)
A1  - Bianchi, Alessandra
A1  - Bregni, Stefano
A1  - Crimaldi, Irene
A1  - Ferrari, Marco
N1  - ?2012 IEEE Global Communications Conference (GLOBECOM 2012)? (3-7 December 2012, Anaheim, California, USA)

Communications QoS, Reliability and Modelling Symposium

N2  - In order to estimate the Hurst parameter of Internet
traffic data, it has been recently proposed a log-regression estimator based on the so-called modified Allan variance (MAVAR). Simulations have shown that this estimator achieves higher accuracy and better confidence when compared with an other method of common use based on wavelet analysis. Here we link it to the wavelets setting and stress why a different analysis for the two approaches is required. We then focus on the asymptotic analysis of the MAVAR log-regression estimator and provide new formulas for the related confidence intervals. By numerical evaluation, we analyze these formulas and make a comparison
between three suitable choices on the regression weights, also optimizing over different choices on the data progression.
SN  - 978-1-4673-0921-9
Y1  - 2012/12//
ER  -