"Multiple testing under unspecified dependency structure"

Two major problems of current multiple testing research are: (i) The "large p, small n" problem, (ii) Appropriately taking into account dependency structures among marginal tests resp. marginal p-values. Our proposed project aims at combining structural assumptions (e.g. block dependencies or pre-definition of hierarchical sub-families of hypotheses like for instance gene ontology terms in genetic studies) and elements from statistical learning theory in order to tackle the practical and methodological challenges.