The Hurst exponent is the simplest numerical summary of self-similarlongrange dependent stochastic processes. We consider the estimation of Hurst exponent in long-range dependent curve time series. Our estimation method begins by constructing an estimate of the long-run covariance function, which we use, via dynamic functional principal component analysis, in estimating the orthonormal functions spanning the dominant sub-space of functional time series. Within the context of functional autoregressive fractionally integrated moving average (ARFIMA) models, we compare finite-sample bias, variance and mean square error among some timeand frequency-domain Hurst exponent estimators and make our recommendations.
- curve process
- dynamic functional principal component analysis
- functional ARFIMA
- long-run covariance
- long-range dependence