If you don't like this particular article, Peter Norvig goes over the same topic (problem?) much more rigorously: http://norvig.com/experiment-design.html