Thumbnail Image

False discovery rate estimation with publication bias

Cao, Tianyu
Recent replication studies have suggested that many pub-lished research findings (discoveries) may be false discoveries in some disciplines. This, in part, explains why results are not replicated in follow-up studies and hence contributes to the “replicability crisis”. However, the breadth and depth of this problem remains largely unknown since costly replication studies are not readily available in most scientific disciplines. This dissertation is aimed at providing tools for estimating the proportion of discoveries that are false discoveries using only published p-values. Consequently, the breadth and depth of the problem can be more readily understood without replication studies.