Fig. 1.

Comparison of IMIX-Ind, IMIX-Cor, IMIX-Cor-Restrict, IMIX-Cor-Twostep, IMIX-AIC, IMIX-BIC, Benjamini–Hochberg false discovery rate (BH-FDR), the Bonferroni correction, the q-value and the local FDR procedure at $\alpha =0.2$. (a) The realized across-data-type false discovery rate in the average of $1\hspace{0.17em}000$ simulations, the results in each scenario are of the average of components 2–8, excluding the global null component 1. (b) Misclassification rate in the average of $1\hspace{0.17em}000$ simulations, the results in each scenario are the average of all components