Gr8wight said:You are so far out in left field, you can no longer see the batter.
example: (using a ridiculously small sample size) Bob receives prayer during your study from a group of 12 people. Tom does not receive any prayer from your study. Do either Bob or Tom receive any prayer from outside your test group? You have absolutely no idea, and no way of finding out without unblinding the experiment by asking them explicitly. Maybe Bob receives no additional prayer, but maybe Tom receives prayer from the entire congregation of his church, 150-200 people. How is Tom's group 'controlled?' It is entirely possible that, acounting for all prayer sources, every member of your test group and of your control group receive exactly the same amount of prayer. But you don't know, because you have absolutely no way of knowing. How can you draw any results from your data if you do not know, can not know, the comparison between how much prayer your test group received compared to how much prayer your control group received?
Even if you attempted to control the study by requesting the friends and family of the control group not to pray for thier loved ones, how successful do you think that would be? How many friends and loved ones, who desired to pray, could be convinced not to. You cannot draw meaningful conclusions from a study when you do not know the values of all the variables. Indeed, when you do not even know what all the variables might be.
Your example is specious in the same way as CFLarsen's.
I control for this with the universally accepted method for designing a controlled experiment: randomization.
If you do not understand the workings of this, I suggest you look at the links I posted earlier, or any elementary statistics or experimental design book.
If you can show that this (majority established) method is flawed, I would of course be delighted to hear your logic.
What you would need to prove, is that an outside factor - any outside variable whatsoever - would be able to cause a difference between control and test groups, when items (trials, people, whatever) have been added to the two groups entirely at random.
If you can do that, you will have invalidated the logic behind every single trial investigating the effects of new medicines in the last couple centuries.
I consider myself to be in the majority scientific concensus on this point, so the burden of proof is on you.
You are of course also welcome to point out how I am improperly applying said methodology in a way that admits a confound.
[Edit:] ... or to claim that randomization is *not* a methodology that is considered sound by a majority of scientists. In which case, I refer you to scholar.google.com - just find a study that investigates the effect of a new drug, which does not use it.