13.7 Benjamini-Hochberg procedure

This procedure is to control FDP deciding which \(H_0\) to reject while \(FDR\approx{E(\frac{V}{R})}\)

\[FDR={E(\frac{V}{R})}<q\]

Less than \(q\), a specific value to use in the algorithm:

ordered p-values: \(p_1<p_2<...<p_m\)

\[L=max\{j:p_{j}<\frac{qj}{m}\}\] reject \(h_{0j}:p_{j}\leq{p_{(L)}}\)

In conclusion, on average no more than a fraction of \(q\) of rejected \(H_0\) are FALSE POSITIVE.

---