Detection of sparse mixtures: higher criticism and scan statistic (Q1711609)

Following \textit{D. Donoho} and \textit{J. Jin} [Ann. Stat. 32, No. 3, 962--994 (2004; Zbl 1092.62051)] and \textit{Yu. I. Ingster} [Math. Methods Stat. 6, No. 1, 47--69 (1997; Zbl 0878.62005)] the problem of detecting a sparse mixture is considered. The main objective of this work is to reveal that scan tests can improve on threshold tests in standard multiple testing settings, particularly when the null distribution has heavy tails. The numerical experiments with \(t\)-distributions with varying numbers of degrees of freedom have demonstrated the theory.