A generalized knockoff procedure for FDR control in structural change detection
Liu, Jingyuan; Sun, Ao; Ke, Yuan
JOURNAL OF ECONOMETRICS Year: 2024 Volume: 239.0
DOI: 10.1016/j.jeconom.2022.07.008
Abstract: Controlling false discovery rate (FDR) is crucial for variable selection, multiple testing, among other signal detection problems. In literature, there is certainly no shortage of FDR control strategies when selecting individual features, but the relevant works for structural change detection, such as pro?le analysis for piecewise constant coe?cients and integration analysis with multiple data sources, are limited. In this paper, we propose a generalized knocko? procedure (GKnocko?) for FDR control under such problem settings. We prove that the GKnocko? possesses pairwise exchangeability, and is capable of controlling the exact FDR under ?nite sample sizes. We further explore GKnocko? under high dimensionality, by ?rst introducing a new screening method to ?lter the high-dimensional potential structural changes. We adopt a data splitting technique to ?rst reduce the dimensionality via screening and then conduct GKnocko? on the re?ned selection set. Furthermore, the powers of proposed methods are systematically studied. Numerical comparisons with other methods show the superior performance of GKnocko?, in terms of both FDR control and power. We also implement the proposed methods to analyze a macroeconomic dataset for detecting changes of driven e?ects of economic development on the secondary industry.
