It's difficult for someone who is self-reporting to do - if all you have is an anecdote. It's not difficult with good experimental design, which is part of my point.