B0152
Title: Local Gaussian autocorrelation and tests of serial independence
Authors: Virginia Lacal Graziani - University of Bergen (Norway) [presenting]
Abstract: The traditional and most used measure for serial dependence in a time series is the autocorrelation function. This measure gives a complete characterization of dependence for a Gaussian time series, but it often fails for nonlinear time series models such as, for instance, the GARCH model, where it is zero for all lags. The autocorrelation function is an example of a global measure of dependence. The purpose is to apply a well-defined local measure of serial dependence, called the local Gaussian autocorrelation, to time series. It generally works well for nonlinear models, and it can distinguish between positive and negative dependence. Asymptotic properties are given and both univariate and bivariate time series are considered. Tests of serial independence based on the local Gaussian correlation are compared to other tests such as the Brownian distance correlation in a number of simulation experiments. The new tests perform well.
