Mistakes in Experimental Design and Interpretation is a nice exposition of the pitfalls you may see when you perform experiments. Remember: any time you measure performance of your code, you are doing an experiment on your system, so the aforementioned caveats apply.
Conan Doyle out-sherlocked by Sherlock (Score:1)
As William Trochim said, Clinical judgement is no substitute for rigorous measurement--and vice versa.
Re: (Score:1)