Estimation of fractal dimension for a class of non-Gaussian stationary processes and fields. (Q1879937)

The authors present asymptotic distribution theory for estimators of the fractal dimension of a random field of the form \(g(X(t))\), where \(g\) is an unknown smooth function and \(X(t)\) is a real-valued stationary Gaussian process or two-parameter Gaussian field whose covariance \(\gamma(t)=\gamma(0)-c| t| ^\alpha+\mathcal{O}(| t| ^{\alpha+\beta})\) obeys the power law at the origin. The variogram satisfies \(\nu(h)\sim 2c| h| ^\alpha\) as \(h\to0\), where the parameter \(\alpha\in(0,2]\) is known as the fractal index. The fractal index can be estimated by fitting a log-regression model to the empirical variogram. It has been observed that the improved estimators can be constructed by considering empirical variograms based on higher-order differences. The authors show that the corresponding limit theory in the non-Gaussian case is somewhat reacher than in the Gaussian case; particularly, because the asymptotic variances may be random. The authors specify the three types of limit distributions that arise if \(g\) is a smooth non-affine function. The first type may be represented as an integral of a certain random function with respect to the Lebesgue measure, the second type is an integral of another random function with respect to an independent Gaussian random measure, and the third type can be represented as a Wiener-Itô integral of order 2. The results of a numerical study are also presented.