Short-Term Noise and the Robustness of Two Log-Periodogram Estimators in Long Memory Series

This paper focuses on the robustness of estimates and its mechanism with presence of short-term noise. Simulation results show that although AG estimator derives lower bias and better robustness than the GPH in most situations, the modification effects are evident only when the short noise has small negative roots. The problem of over-modification on larger negative roots and the under-modification on the positive roots are still lack of advanced study. The standard deviation it is not sensitive to short-term noise but the mean square errors increase sharply with short-term noise. Besides, the power and practical size of the test was affected too. Larger sample size is suggested to gain more robust finite sample properties.