Confronting classical and Bayesian confidence limits to examples