I've come across a very well-written and incisive PDF—sadly anonymous—which I recommend to anyone interested in the use and misuse of statistics. It's a good complement to the work of John Brignell, who wrote two excellent books about the vanity of ultra-low risk epidemiologists about ten years ago. It focuses on lung cancer risk and it has the most comprehensive annotated list of passive smoking studies yet compiled (more so than the one I produced in 2008, which was limited to female exposure).
Don't be put off by the title ('The Plain Truth About Tobacco'). It's not about plain packaging and it doesn't argue that smoking is not the main cause of lung cancer (in fact, any remaining deniers should be forced to read it), but it does give a well-deserved kicking to those who cite statistics without bothering to understand what they mean.
You would not believe the things that epidemiologists have tried to "link" lung cancer to. If you've ever heard that you can get lung cancer (or protect yourself from lung cancer) by owning a pet canary, see pages 92 to 106 of this book. I learnt a lot and I think you will too.
Download it here. Stick it on your iPad or Kindle or whatever and away you go...