he False Discovery Rate (FDR) was proposed in Benjamini and Hochberg (1995) as a powerful approach to the multiplicity problem that does not require strong control of the familywise error rate (FWER). The original approach was developed for independent test statistics and was lately extended to generally dependent statistics in Yekutieli (2008). In this paper we extend the existing results by showing that specific copulæ can be used to represent the dependence structure among classes of univariate statistics that leads to FDR control. In particular, within a fairly general setting we use Liebscher (2008, 2011) and Muller and Scarsini (2005) and show that FDR control is ensured when the dependence existing among the statistics is described by asymmetric Archimedean copulae. In deriving our results we extend Yekutieli (2008).
A copula-based analysis of false discovery rate control under dependence assumptions
LUPI, Claudio
2012-01-01
Abstract
he False Discovery Rate (FDR) was proposed in Benjamini and Hochberg (1995) as a powerful approach to the multiplicity problem that does not require strong control of the familywise error rate (FWER). The original approach was developed for independent test statistics and was lately extended to generally dependent statistics in Yekutieli (2008). In this paper we extend the existing results by showing that specific copulæ can be used to represent the dependence structure among classes of univariate statistics that leads to FDR control. In particular, within a fairly general setting we use Liebscher (2008, 2011) and Muller and Scarsini (2005) and show that FDR control is ensured when the dependence existing among the statistics is described by asymmetric Archimedean copulae. In deriving our results we extend Yekutieli (2008).I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
