A PP asked why people would discount anecdotal evidence. I think it's because it is, by nature, unreliable (in a scientific sense) due to the small sample sizes.
If you were told (for example) that out of 1 million people doing X all were recorded, under test conditions, as having Y response then you would feel pretty certain that X led to Y. Even if 1 person in that study ended up with Z, you would still say that if you were doing X you were pretty much definitely going to get Y.
Now if someone told you that despite that evidence, they had heard of a few people who had experienced Z, and so the study was replicated across different countries many many times, but all found that between 999,999 - 1 million people all got Y in every study...
Well. In my case I would think that I believe X leads to Y for such a majority of cases to render anything else statistically insignificant. And by insignificant I don't mean that Z (in this case) would be insignificant, but I would say that Z is insignificant in terms of its linkage to X.
It's a bit of a laboured analogy, but I hope it makes sense.
There are many things that 'feel' true, but sometimes are really not. Correlation does not equal causation.
If you have studied behavioural science then you may know of Skinner's experiements with pigeons, whereby they often end up doing what he termed 'superstitious dances' to get food, simply because they were doing a particular thing at a time when food was randomly delivered, so they repeated it to try to get the food again.
I am not trying to insult anyone - just to explain why, for me at least, science prevails over anecdotal evidence.