Remote healing

[Zep]

I have read your protocol, the holes (nay, chasms) in it have been explained to you in detail ad infinitum, and you are now definitely sounding like a record stuck in one groove. You might care to rein in your angst somewhat, and have an actual read of what the other scientists in this thread have told you - they do this sort of thing for a living, and so know a thing or two about it. Belligerently and unbendingly pushing your protocol while not taking any freely available sage advice on board is not wise science, and hardly likely to lead to a positive outcome for you in any respect.

But don't let us stop you.

Good luck, and goodbye from me - I have better things to do here.

 

[saizai]

I have read your protocol, the holes (nay, chasms) in it have been explained to you in detail ad infinitum, and you are now definitely sounding like a record stuck in one groove. You might care to rein in your angst somewhat, and have an actual read of what the other scientists in this thread have told you - they do this sort of thing for a living, and so know a thing or two about it. Belligerently and unbendingly pushing your protocol while not taking any freely available sage advice on board is not wise science, and hardly likely to lead to a positive outcome for you in any respect.

But don't let us stop you.

Good luck, and goodbye from me - I have better things to do here.

You haven't replied to my question.

Please explain, exactly, how it is that randomization does not control for all uncontrolled variables, including specifically those you have brought up.

Please. I'm really honestly curious how you would make this argument.

 

[CFLarsen]

saizai,

How about those calculations?

 

[Pup]

I also think it [only one prayer works; the rest are superfluous] is completely impossible to test for as a theological argument, and akin to other apologism that asserts untestability - ignorable as a matter of faith alone. You are welcome to believe whatever you like on it.

Do you have any suggestion to how this is untrue? E.g., do you have any way to test for it?

Well, theoretically and somewhat facetiously, yes. An experiment could be set up in which something so specific is prayed for that the chances of it being prayed for randomly are almost zero. Half the subjects have someone pray that a person carrying a chocolate pie walks into their hospital room, and half don't. But then of course there are all sorts of objections about sincerity and trivialization of prayer.

I am not "convinced". As I have said - I am agnostic on this issue, and doing this out of curiousity.

Perhaps "convinced" isn't the right word, but I mean something like "convinced a specific hypothesis is worth testing." Ironically, it's easier to design a test for a whacko claim that some nutball has convinced himself is true, like "rubbing applejuice on your head will make a bald person grow hair within a week" than to design a test for something reasonable but vague, like "applejuice is good for you."

So it's almost impossible to design a single test for a vague claim that "prayer or remote healing works." It just seems to me that your test is designed to test if prayer/healing works in a very limited and specific way, yet it seems that you're not strongly motivated to think that the specific hypothesis you're testing for is the most likely one. Thus there's a strong temptation for data mining or other ways of finding false positives in unexpected ways.

 

[SezMe]

I don't think JREF is going to accept this - or any similar - challenge on religious grounds. But IF they did, one of the first changes that they will require is that you do some rigorous statistics that back up the protocol and define success criteria.

In other words, the "numbers" Larsen keeps ragging you about. So do 'em now or do 'em later. It'd be better to do the numerical analysis now so we can avoid having to cite Larsen's post count in logrithmic terms.

 

[saizai]

I don't think JREF is going to accept this - or any similar - challenge on religious grounds. But IF they did, one of the first changes that they will require is that you do some rigorous statistics that back up the protocol and define success criteria.

In other words, the "numbers" Larsen keeps ragging you about. So do 'em now or do 'em later. It'd be better to do the numerical analysis now so we can avoid having to cite Larsen's post count in logrithmic terms.

Larsen's "numbers" is, well, trolling. The argument for the radnomization method is a logical one, and very very quickly goes axiomatic - i.e. it rests on the definitions of randomness and causality. Larsen has yet to address the basic argument I gave in any way, nor to address why zie doubts the validity of a centuries-established and very viable scientific method. Nor to take up my challenges - e.g. to find a study re. a medicine's effectiveness that does not use it.

I'll start a new thread for this.

Please point out what other protocol criteria or statistics are in any way unclear? I believe I have defined everything...

 

[saizai]

Well, theoretically and somewhat facetiously, yes. An experiment could be set up in which something so specific is prayed for that the chances of it being prayed for randomly are almost zero. Half the subjects have someone pray that a person carrying a chocolate pie walks into their hospital room, and half don't. But then of course there are all sorts of objections about sincerity and trivialization of prayer.

Interesting idea. You would still need something that *does* occur, though - harder to show an effect that large, so it's a bit unfair against. But you could certainly pray for something highly unusual to pray for but commonplace to happen; I'm sure you can think of any number of such things; ideally you'd want something that happens about half the time. I think PEAR is one form of this; I know nothing about its methodology or potential flaws thereof, though, so I make no claim as to whether it does it well.

Perhaps "convinced" isn't the right word, but I mean something like "convinced a specific hypothesis is worth testing." Ironically, it's easier to design a test for a whacko claim that some nutball has convinced himself is true, like "rubbing applejuice on your head will make a bald person grow hair within a week" than to design a test for something reasonable but vague, like "applejuice is good for you."

So it's almost impossible to design a single test for a vague claim that "prayer or remote healing works." It just seems to me that your test is designed to test if prayer/healing works in a very limited and specific way, yet it seems that you're not strongly motivated to think that the specific hypothesis you're testing for is the most likely one. Thus there's a strong temptation for data mining or other ways of finding false positives in unexpected ways. [/B]

Rephrase your last paragraph?

As for data mining - the two-step process removes any possibility of this being a confound in the final result.

 

[eri]

I understand your point about randomization, and I see that you probably mean that the prayer being done for any of these patients at any given time by someone not participating in the subject will simply serve as background noise in the data. But I don't think you can make a case for your specific prayer being done standing out in any way. Like Pup said, we don't have a 'basic unit' of prayer to go by here; we don't know if prayer is additive. For instance, many people think that the higher up someone is in the church (say, the pope), the more effect their prayer might have. It's just that even if one prayer = one prayer unit (PU), the effect of anyone that might know that person praying for them (and the people who routinely pray for the whole world anyway) would far out number (by orders of magnitude - you could probably safely assume 100 PU/day minimum) your added effect of a professional prayer (1 PU). That extra prayer would be completely swamped by the others; you could never make a statistical case for separating it. I'm afraid there are just too many factors here to ever argue for a statisically significant result. It's a good idea, and I'm sure we'd all love to see a test that could actually test the effects of prayer, but this way has far too many uncontrollable factors.

But if you were to conduct it, what statistical test did you have in mind? I suppose you could start with a simple truth table. You would have to quantify 'improvement', however, which may be difficult if your subjects have more than one illness among them. In order to have statistically significant results, you are going to need a large test pool as well, which will add to the complications already in place.

 

[saizai]

I understand your point about randomization, and I see that you probably mean that the prayer being done for any of these patients at any given time by someone not participating in the subject will simply serve as background noise in the data. But I don't think you can make a case for your specific prayer being done standing out in any way. Like Pup said, we don't have a 'basic unit' of prayer to go by here; we don't know if prayer is additive. For instance, many people think that the higher up someone is in the church (say, the pope), the more effect their prayer might have. It's just that even if one prayer = one prayer unit (PU), the effect of anyone that might know that person praying for them (and the people who routinely pray for the whole world anyway) would far out number (by orders of magnitude - you could probably safely assume 100 PU/day minimum) your added effect of a professional prayer (1 PU). That extra prayer would be completely swamped by the others; you could never make a statistical case for separating it. I'm afraid there are just too many factors here to ever argue for a statisically significant result. It's a good idea, and I'm sure we'd all love to see a test that could actually test the effects of prayer, but this way has far too many uncontrollable factors.

Again, these are theological speculations.

I think you have already agreed that everything other than what's done in the study goes under "background noise".

You can certainly make an argument that what the study does is *not sufficient* or *moot* according to your theoelogical perspective - in which case you would be arguing that a negative result would not disprove the existence of prayer as you describe it. I'm OK with that.

But if you were to conduct it, what statistical test did you have in mind? I suppose you could start with a simple truth table. You would have to quantify 'improvement', however, which may be difficult if your subjects have more than one illness among them. In order to have statistically significant results, you are going to need a large test pool as well, which will add to the complications already in place.

I descrbied that - the measure is numeric. Two groups each have measures. (Did you read the full study design post - version 2?)

Compare the two for statistical significance of difference. QED.

 

[eri]

I have read the post (second version) again at your request. I still don't see exactly how you plan to do the statistics. How do you plan to compare the results to each other? Graphically? Using valid statistical tests? How exactly will you get to your statistical measure of significance? The 'comparison of one to the other' isn't good enough, I would like to see test determined ahead of time that will conclusively determine if you have detected a difference.

I was trying to point out that sometimes the signal is just too low to be subtracted from the background. I think your signal-to-noise ration will be miniscule, and I don't know if you will be able to argue that it's there or not. I seriously doubt it the signal will be seperable.

 

[saizai]

I was trying to point out that sometimes the signal is just too low to be subtracted from the background. I think your signal-to-noise ration will be miniscule, and I don't know if you will be able to argue that it's there or not. I seriously doubt it the signal will be seperable.

This is possible. But, in any case, you would also have the converse result - i.e. X% certainty that there is *no* effect, greater than Y%. Of course, your sample size will dictate how fine a result you can detect to a decent certainty.

 

[Gr8wight]

Thank you, Pup, for explaining that in a way that a layman like me can understand. I see exactly what is being said now. It seems to me that, in order to assume that the control group and the test group are equal before the 'extra' prayer is added, a huge sample size would be required, in the order of several hundred participants at the smallest. Saizai is not talking about the one million dollar challenge, but a clinical study. One would wonder exactly what he hopes to gain by posting here.

I've been reading this thread, and I'm puzzled at the point being argued. It seems to me Sai's randomization *does* control for others praying. The test is so narrow as to be almost worthless (see, I'm not really on his side ), but it's a valid test for what it does.

It's testing to see if a little extra prayer makes any difference. Not if prayer makes a difference, period.

One could compare it to testing for anything that can't be reduced to zero. Let's say, noise. Or exercise. Or sunlight.

Everyone is exposed to a certain amount of sound or sunlight or experiences a certain amount of movement each day. You can't reduce it to zero, by telling the control group not to move or never to go outside or to wear earplugs for a month, but by picking two random groups that are large enough, it's reasonable to assume that the "background exposure" is going to average out the same for both.

So then you add the variable. In addition to their normal activities, one group has to be exposed to 20 minutes of loud noise every other day, or has to walk on a treadmill for half an hour three times a week, or has to be in the sun for an hour each day, or whatever. You'd need to make sure the variable didn't make them change their normal activities to compensate--almost impossible in my hypothetical examples--but with double-blind prayer, that wouldn't be a problem.

So now you've got a control group getting X amount of whatever you're testing for, due to normal "background exposure," and another group getting X+1. It seems a valid way to test whether getting X amount of prayer +1 makes people heal better than just getting X amount of prayer.

Which is a test so narrow that it hardly seems useful. The example of the Pope (or any famous person) who gets more prayers than an unknown sufferer, indicates that extra prayers apparently don't help, since famous people appear to suffer the same ills as non-famous people. And as someone else said, if X is large and 1 is small, it would be like trying to hear a whisper in a football stadium.

But I honestly don't see the objection to using randomization to equalize the amount of "background prayer" in both groups. Maybe I'm missing something.

 

[saizai]

Thank you, Pup, for explaining that in a way that a layman like me can understand. I see exactly what is being said now. It seems to me that, in order to assume that the control group and the test group are equal before the 'extra' prayer is added, a huge sample size would be required, in the order of several hundred participants at the smallest. Saizai is not talking about the one million dollar challenge, but a clinical study. One would wonder exactly what he hopes to gain by posting here.

Huge sample size: Not really. It mainly depends on two things - what your required certainty is (95% is the standard), and how small of an effect you want to be able to detect. The smaller, the more people - exponentially.

I am talking about both a clinical study and the challenge. Given its topic, a positive result would constitute what the challenge requires: a statistically significant, verifiable demonstration of something that cannot be explained by (and in fact is denied as being possible at all by) mundane science in its current form.

I realize that this is not the usual form of such challenges, since I am not trying to demonstrate any particular power of my own that occurs immediately, but it meets all the requirements.

 

[Ashles]

Excellent work Sai.

I have been following this post very closely, and commend you for your clarity, and also your patience with those (few) openly rude posters.

Though I am no statistics boffin, it seems that your proposal is pretty sound, and these attempts by you to firm up the method can only improve it.

Thank you for your efforts so far.

I hate to appear cynical here, but we have previously had a poster here who has also vocally supported his own opinion with sockpuppets with very few posts of their own.

That posters name? Lucianarchy.

Also, Lucianarchy really loves to criticise CF Larsen at any available opportunity.

Now I'm not saying that saizai is Luci...

I'm just saying let's see.

 

[saizai]

I hate to appear cynical here, but we have previously had a poster here who has also vocally supported his own opinion with sockpuppets with very few posts of their own.

That posters name? Lucianarchy.

Also, Lucianarchy really loves to criticise CF Larsen at any available opportunity.

Now I'm not saying that saizai is Luci...

I'm just saying let's see.

a) "Boffin" is British English. You'll note that my accent is Californian geek.

b) Ask an admin to check IPs. I make no claims for others, but this is my only account on this board.

 

[Gr8wight]

Huge sample size: Not really. It mainly depends on two things - what your required certainty is (95% is the standard), and how small of an effect you want to be able to detect. The smaller, the more people - exponentially.

I am talking about both a clinical study and the challenge. Given its topic, a positive result would constitute what the challenge requires: a statistically significant, verifiable demonstration of something that cannot be explained by (and in fact is denied as being possible at all by) mundane science in its current form.

I realize that this is not the usual form of such challenges, since I am not trying to demonstrate any particular power of my own that occurs immediately, but it meets all the requirements.

What sample size do you propose? As well, who is going to fund your study?

 

[Flange Desire]

a) "Boffin" is British English. You'll note that my accent is Californian geek.

b) Ask an admin to check IPs. I make no claims for others, but this is my only account on this board.

Hmmmmmm
Just did a search for "boffin", and the only occurrance I could find was an earlier post of mine in this thread.
I am not a sockpuppet! (echos from the movie The Elephant Man).
Yes, we use a bit of "British English" down here in oz.

 

[saizai]

What sample size do you propose? As well, who is going to fund your study?

As a preliminary, 200-600; my current design says 400, but that could be changed. That should give enough to capture an effect from neutral of ~6% at p<.05 (numbers from a math-major friend of mine).

The secondary / full formal, if the preliminary is successful, would probably try to have at least a few thousand. Funding, and participants, should not be hard to obtain if (if) I managed to be the only person ever to pass the preliminary.

Funding - *shrug* I don't have any. I doubt I could get my university to fund it, but I could try. I would probably have to run entirely on donations / volunteers - all participants volunteer, get someone to donate server space, run as much as possible online, or via having them snailmail me, etc. Perhaps a religious org would be willing to help fund such a thing. I consider that an implementation detail, one that I'll address as needed.

 

[Moochie]

What is prayer, exactly?

Hi,

I've just completed reading a portion of this thread and am puzzled as to what everyone thinks prayer actually is.

As a lay person, I have seen nothing here that convinces me that, regardless of the experimental design and sample size, one could in any way prove cause and effect.

I seriously question the designer's motives in attempting to prove the efficacy of "prayer."

Regards,

Moochie

 

[saizai]

Re: What is prayer, exactly?

As a lay person, I have seen nothing here that convinces me that, regardless of the experimental design and sample size, one could in any way prove cause and effect.

Explain how this would not demonstrate cause and effect, assuming there is an effect?

I seriously question the designer's motives in attempting to prove the efficacy of "prayer."

Pray, tell me what your questions are regarding my motives?