出版社：Suntory Toyota International Centres for Economics and Related Disciplines
摘要：The aggregation procedure when a sample of length N is divided intoblocks of length m = o(N), m →∞and observations in each block arereplaced by their sample mean, is widely used in statistical inference.Taqqu, Teverovsky and Willinger (1995), Teverovsky and Taqqu (1997)introduced an aggregate variance estimator of the long memoryparameter of a stationary sequence with long range dependence andstudied its empirial performance. With respect to autovariance structureand marginal distribution, the aggregated series is closer to Gaussianfractional noise than the initial series. However, the variance typeestimator based on aggregated data is seriously biased. A refinedestimator, which employs least squares regression across varying levelsof aggregation, has much smaller bias, permitting derivation of limitingdistributional properties of suitably centered estimates, as well as of aminimum mean squared error choice of bandwidth m. The results varyconsiderably with the actual value of the memory parameter
关键词：Long memory; aggregation; semiparametric model
