Statistical analysis of large dimensional long-memory time series and its applications

2018 Fiscal Year Final Research Report

• Project/Area Number: 15K17038
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Economic statistics
• Research Institution: Okayama University
• Principal Investigator: Narukawa Masaki 岡山大学, 社会文化科学研究科, 准教授 (30588489)
• Project Period (FY): 2015-04-01 – 2019-03-31
• Keywords: 長期記憶 / 高次元時系列 / セミパラメトリック法 / 因子モデル / Taper

Outline of Final Research Achievements

This research developed semiparametric statistical analysis of large dimensional time series with long-range dependence by using extended factor models. Specifically, we proposed the two-step semiparametric approach in which the common components are estimated by principal components analysis as the first step and the estimators of the memory parameters are obtained by the local Whittle method as the second step. We also investigated the asymptotic properties and the finite sample performance. Moreover, we constructed multivariate local Whittle estimators by incorporating the maximal efficient taper, and provided an empirical application of the proposed method to exchange rate data.

Free Research Field

計量経済学

Academic Significance and Societal Importance of the Research Achievements

次元の大きい長期記憶性を持つ時系列に対しては，既存分析手法の適用が困難な上に関連研究の蓄積が乏しい中，本研究で考案している時間領域の因子モデルと主成分分析法に周波数領域のセミパラメトリック推測を組み合わせた2段階アプローチは，二つの領域の手法を駆使することで高次元長期記憶時系列における統計的推測法を提供しうるという大きな意義があり，時系列データの統計的分析に新たな視点と方向性をもたらすであろう．