Lies, damned lies and meaningless correlations
While I was researching for a review of Standard Deviation, a book by Gary Smith which explains how statistics can be used - or misused - to support more or less any argument, I came across this website devoted to spurious correlations.
The website's author, Tyler Vigen, wrote a program to find correlations between completely unrelated sets of data. The resulting charts are hilarious. This one "correlating" the number of people who died tangled in their bedsheets with the per capita consumption of cheese.
Smith's book is a great primer if you don't want to be baffled by "science". I particularly like the quote from Nobel Prize winning economist Ronald Coase, who said: “If you torture data long enough, it will confess.”