If F is the number of false discoveries of some multiple testing procedure, then show the per-family error rate E(F) satisfies the crude inequalities P{F ≥ 1} ≤ E(F) ≤ s P{F ≥ 1} , where s is the number of hypotheses under test. Hence, if a method controls E(F)

at α, then it also controls the FWER at α. Show that the Bonferroni method controls both at α. What about the Holm method?