2018/09/05 by Eugene Katsevich, Katsevich, Eugene, Chiara Sabatti +3 · 1 citation
Biochemistry, Genetics and Molecular Biology · #Bioinformatics and Genomic Networks #FOS: Computer and information sciences #Gene expression and cancer classification #Genetic Associations and Epidemiology #Methodology (stat.ME)
paper · pdf · doi:10.48550/arxiv.1809.01792
openalex publication_date 2018/09/05 · openalex created_date 2022/08/03 · openalex updated_date 2026/07/28
Scientific hypotheses in a variety of applications have domain-specific\nstructures, such as the tree structure of the International Classification of\nDiseases (ICD), the directed acyclic graph structure of the Gene Ontology (GO),\nor the spatial structure in genome-wide association studies. In the context of\nmultiple testing, the resulting relationships among hypotheses can create\nredundancies among rejections that hinder interpretability. This leads to the\npractice of filtering rejection sets obtained from multiple testing procedures,\nwhich may in turn invalidate their inferential guarantees. We propose Focused\nBH, a simple, flexible, and principled methodology to adjust for the\napplication of any pre-specified filter. We prove that Focused BH controls the\nfalse discovery rate under various conditions, including when the filter\nsatisfies an intuitive monotonicity property and the p-values are positively\ndependent. We demonstrate in simulations that Focused BH performs well across a\nvariety of settings, and illustrate this method's practical utility via\nanalyses of real datasets based on ICD and GO.\n