Figure 7
From: A comparative review of estimates of the proportion unchanged genes and the false discovery rate
FDR for the Golub et al. [32] data. The dotted line represents a cut-off that minimises the number of false positives and false negatives, summing (4) and (5) using the smoother estimate. It would make sense to consider choosing a threshold in this region. The line corresponds to the 2920th ordered p-value and a q-value of just above 0.2. All methods agree rather well. For reference the local FDR of [29] has been added. As expected it exceeds all others, and is caught up at zero by its FDR counterpart SEP FDR. However, they meet at a level well above other methods.