.... 1 bottle out of 10 is homeopathic remedy, and 9 bottles are pure water. Each person is given a bottle at random, and they have to correctly identify if it is water or remedy.
Now, it's no longer 50/50, so does 3 out of 6 sound impressive? Considering if all people simply answered "water", we would expect 5, maybe even 6, to be right.
No, wait a minute, I see the germ of a good idea here.
The reason they say they can't identify which one of a number of remedies they've taken is that to get the full symptom picture you need a team of at least ten provers.
So, we change the test to a simple matter of distinguishing a remedy chosen by the homoeopath for its ability to produce striking symptoms (however non-specific) in one person, and ask them to say if they've been given that, or the blank pills. A simple either/or test here produces problems because obviously there is a 50:50 chance of a lucky guess.
However, what if we were to give the homoeopath
ten or more vials, with the assurance that one and
one only contained the specified remedy. Surely they should be able to tell which one quite easily! Now as described that still leaves a 1 in 10 chance of a lucky guess, which is still too much. However, it's a lot better than 1 in 50, as regards working towards statistical significance.
So, we need to find out how many vials a homoeopath thinks it would be practical to screen for remedy presence, assuming only one is the actual remedy. Then we need to work out how many repetitions of that test would require to be correct in order to achieve p<0.001. If in fact this level of significance can be achieved with perhaps only three repetitions, then it could well be practical for one person to undertake the whole thing, just saying when they felt ready to go on to the next set. Even if it was a couple of months, that would be achievable.
Now, what excuses might they find for not doing this one?
Rolfe.