A multi-step procedure to determine the number of factors in large approximate factor models. Issue 17 (9th August 2021)

• Record Type: Journal Article
• Authors: Luo, Ronghua; Jiang, Jiakun; Lan, Wei; Yan, Chengliang; Ding, Yue
• Is Part Of: Communications in statistics. Volume 50:Issue 17(2021)
• Page Start: 3988
• Page End: 3999
• Publication Date: 2021-08-09
• Subjects: Approximate Factor Models -- eigenvalue ratio criterion -- multi-step procedure -- factor number; Mathematical statistics -- Periodicals; Mathematics; Statistics; 519.2
• DOI: 10.1080/03610926.2019.1710752
• Languages: English
• ISSNs: 0361-0926

Abstract

Abstract: We propose in this article a multi-step procedure to determine the number of common factors in large approximate factor models (Chamberlain and Rothschild 1983 ) in the spirit of the eigenvalue ratio (ER) criterion of Ahn and Horenstein (2013 ). We show theoretically that this multi-step procedure can consistently identify the true number of factors while both relatively strong and weak factors simultaneously exist, which is not easy to perform with the ER criterion. Our extensive simulation results demonstrate that the proposed procedure has better finite sample properties compared with that of the ER criterion under certain cases.