[dkersten]

The problem with anecdotes is that we don't actually know what the cause of the results was. Maybe its the thing in question, but it could be something else either. That's the whole point of control groups (so you have a baseline to compare against and can filter out "contamination") and double blind tests (so you remove bias).

If you have suspect data, throwing more suspect data in the pile doesn't make it less suspect. Anecdotal evidence is just that: unproven possibly suspect data. Having more of these doesn't make it any more reliable.

So regardless of the original quote, I stand by the plural of anecdote not being data.

[rrherr]

So your concern isn't just with anecdotes, but with observational data in general?

I agree that experimental data is preferred, but sometimes observational data is all we've got, right?

[dkersten]

Sure and it can be a good reason to go "oh, that was interesting! Let's study it!"

It is not, however, good enough to treat as fact or make any general inferences from because it's not very strong evidence.