Expectiles for subordinated Gaussian processes with applications (Q1950818)

It is well known that the increment process of fractional Brownian motion (fBm) has a correlation function with asymptotic power law decay \(\rho_H(i)\sim|i|^{2H-2}\), where \(H\in(0,1)\) is the Hurst exponent. More generally, the authors study subordinated stationary Gaussian processes with unit variance and correlation function of asymptotic power law decay. A Bahadur-type representation for sample expectiles of such processes is presented which proves asymptotic normality of the sample expectiles in case of short-range dependence. This allows to define new discrete variation estimators for the Hurst exponent of fBm which rely on scale and location equivariance properties of the expectiles. The estimators are refinements of methods by the first author by substituting sample variance, respectively quantiles, by expectiles. Further, consistency and asymptotic normality of the estimators are proved. The authors finally provide a simulation study in order to confirm the effectiveness of their estimation procedures.