Conventional works to estimate initial parameters for multivariate monofractal time series are not available for multifractal time series. This paper deals with the initial parameter estimation for multi-variate multifractal time series based on the multifractal time transformation. At first, we summarize the parameter estimation scheme of multivariate monofractal time series. Then, we introduce multifractal time transformation to extend the estimation scheme. In the method we only use the exponent r(q) obtained from q-th moment of wavelet coefficients for scaling T(q) rather than the transformation itself. In the estimation by using the relations obtained from wavelet coefficients of time series, we simplify the estimation based on the relation of Hausdorff dimension of time series. As applications, we examine the estimation of our methods based on the generation of multivariate multifractal time series by using the MWM (Multifractal Wavelet Model).
展开▼
