You should not rely only on empiric methods, but as well on analysis. If you have a change of the situation, that from other analysis makes a lot of sense, that it will have an effect, an outlier is something quite different.
Or in other words. Frequentist methods sucks, compared to Baysian approaches, that figure in a priori probability densities. I take dodo's graph with the explaination any time as a evidence that the methods proposed in Hungary work.

Nevertheless:
Freie Fahrt für freie Bürger ;-p

Der Amerikaner ist die Orchidee unter den Menschen
Volker Pispers