Figure 2

Expression dataset gene set analysis, smokers versus never-smokers. Gene set analyses were performed by the (A) hypergeometric and (B) Wilcoxon rank tests using gene sets defined by the Gene Ontology, and the replication of each gene set was assessed via our gene set bagging procedure (each point is one gene set). The relationship between the estimated replication probability ($\widehat{R}$) and traditionally reported p-value appears much more concordant using the Wilxocon rank test.