rationalskepticism.org

I suppose it is redundant to state the obvious but it seems to get missed so often:

Sort the data that is relevant to the problem from the data that is not.

psik