Title:Kolmogorov-Smirnov Estimation of Self-Similarity in Long-Range Dependent Fractional Processes

Abstract:This paper investigates the estimation of the self-similarity parameter in fractional processes. We re-examine the Kolmogorov-Smirnov (KS) test as a distribution-based method for assessing self-similarity, emphasizing its robustness and independence from specific probability distributions. Despite these advantages, the KS test encounters significant challenges when applied to fractional processes, primarily due to intrinsic data dependencies that induce both intradependent and interdependent effects. To address these limitations, we propose a novel method based on random permutation theory, which effectively removes autocorrelations while preserving the self-similarity structure of the process. Simulation results validate the robustness of the proposed approach, demonstrating its effectiveness in providing reliable estimation in the presence of strong dependencies. These findings establish a statistically rigorous framework for self-similarity analysis in fractional processes, with potential applications across various scientific domains.