A1193
Title: Testing block-diagonal covariance structure under a low dimensional factor model in high-dimensional settings
Authors: Takahiro Nishiyama - Senshu University (Japan) [presenting]
Masashi Hyodo - Kanagawa University (Japan)
Shoichi Narita - Kanagawa University (Japan)
Abstract: The aim is to propose a new test for block-diagonal covariance structure in a high-dimensional framework, while accommodating a low-dimensional latent factor model. The test, built under low-dimensional factor models, distinguishes from previous normal approximation-based tests, which are valid under a weak spike structure. A modified RV coefficient is proposed for high-dimensional data, and it is shown that its null-limiting distributions follow a weighted mixture of chi-square distributions under a high-dimensional asymptotic regime integrated with weak technical conditions. By applying this asymptotic result and estimation theory of the number of factors in a low-dimensional factor model, a new approximation test is proposed for a block-diagonal covariance structure. The finite sample and dimensional performance of this test are also examined using Monte Carlo simulations.
