Frequentist versus Bayesian approaches to multiple testing