"Lies, damned lies and statistics."

Someone once said to me that it was easier to discuss statistics with someone who is ignorant of statistics rather than someone who knows a little bit about it, as the person who has some knowledge will probably get it wrong. This is largely, I suspect, because it's very easy to overlook assumptions and seemingly minor bits of information which turn out to be extremely important.

This is a 20 minute TEDtalk from 2005 which I quite like which illustrates that point.

PS. If you've never been to TED, you ought to. There are plenty of really interesting talks there.